2023 · paper
Coarse-to-Fine Multi-Scene Pose Regression With Transformers
The institutional record describes a transformer architecture for multi-scene camera localization evaluated on indoor and outdoor benchmarks.[128]
deep learning, uncertainty and multimodal learning for complex data
Yoli Shavit
יולי שביט
Identity: verifiedA July 2025 Bar-Ilan announcement and institutional publication record identify Yoli Shavit and match the roster spelling; the announcement describes a lab planned for the 2025-2026 academic year, not independently verified current operations. [127][128]
Documented foundation
Research & experience
Bar-Ilan announced in July 2025 that Shavit would lead a deep-learning laboratory focused on generalization, uncertainty, multimodality and complex data, including biological applications; a separate institutional record verifies her transformer camera-localization work.[127][128]
uncertainty-aware deep learningmultimodal representation learningcamera localizationcomplex biological data modelling
CV and official profile
Official laboratory and publication pages were verified, but no standalone current CV was found in the bounded pass.[127][128]
Selected work
Representative records, not a complete publication list. Metadata confirms attribution; it does not independently replicate a result.
Patent evidence
5 catalogued patent records · 5 identified families
Coverage: Partial inventor search
Seven individual patent documents from the public inventor index were inspected, grouped into five recorded families. Two log-encoding publications share family 97303957; the two job-failure applications share family 98777322. NVIDIA conference affiliation and SemantiLog work support the patent team/topic mapping; these are Mellanox/NVIDIA records, not an affiliation-filtered Bar-Ilan inventory. Post-cutoff or unindexed records are not claimed.
Multi-modal attribution for job failure in a distributed system
US20260050507A1 · Published 2026-02-19
Published patent document inspected
Publication assignee: Nvidia Corp
The inspected inventor field names Yoli Shavit. The ASE 2024 first-party record identifies Yoli Shavit at NVIDIA working on SemantiLog. The patent subject and recurring Shteingart/Zahavi/Mataev team align; the faculty announcement establishes the Bar-Ilan researcher. Same inspected family: US20260050506A1. One representative retained; variants intentionally grouped. Assignee means the observed original-assignee field; no current ownership conclusion.[328][329][426][427]
Using neural networks to encode log data
US12499303B2 · Published 2025-12-16
Published patent document inspected
Publication assignee: Mellanox Technologies Ltd
The inspected inventor field names Yoli Shavit. The ASE 2024 first-party record identifies Yoli Shavit at NVIDIA working on SemantiLog. The patent subject and recurring Shteingart/Zahavi/Mataev team align; the faculty announcement establishes the Bar-Ilan researcher. Same inspected family: US20260064938A1. One representative retained; variants intentionally grouped. Assignee means the observed original-assignee field; no current ownership conclusion.[302][330][426][427]
Using similarity loss to train neural networks
US20250335762A1 · Published 2025-10-30
Published patent document inspected
Publication assignee: Mellanox Technologies Ltd
The inspected inventor field names Yoli Shavit. The ASE 2024 first-party record identifies Yoli Shavit at NVIDIA working on SemantiLog. The patent subject and recurring Shteingart/Zahavi/Mataev team align; the faculty announcement establishes the Bar-Ilan researcher. Assignee means the observed original-assignee field; no current ownership conclusion.[325][426][427]
Using contrastive learning to train neural networks
US20250335761A1 · Published 2025-10-30
Published patent document inspected
Publication assignee: Mellanox Technologies Ltd
The inspected inventor field names Yoli Shavit. The ASE 2024 first-party record identifies Yoli Shavit at NVIDIA working on SemantiLog. The patent subject and recurring Shteingart/Zahavi/Mataev team align; the faculty announcement establishes the Bar-Ilan researcher. Assignee means the observed original-assignee field; no current ownership conclusion.[324][426][427]
Using neural networks to classify logs
US20250335549A1 · Published 2025-10-30
Published patent document inspected
Publication assignee: Mellanox Technologies Ltd
The inspected inventor field names Yoli Shavit. The ASE 2024 first-party record identifies Yoli Shavit at NVIDIA working on SemantiLog. The patent subject and recurring Shteingart/Zahavi/Mataev team align; the faculty announcement establishes the Bar-Ilan researcher. Assignee means the observed original-assignee field; no current ownership conclusion.[323][426][427]
Original report snapshot
Original evidence: not verified
No attributable patent record was verified in the bounded search.
Records are counted separately from identified families. Author-reported entries are labelled and may still need publication verification. Inventorship, publication-time applicant and current ownership are different facts. No legal-status, patentability or freedom-to-operate conclusion is made.
Scores prioritize research fit from 1–10; they are not probabilities.
Review: Reviewed with limitations
Proposed capability matches, not confirmed relationships. Scores are analyst judgments with low forecast confidence; researcher interests, capacity and feasibility need confirmation.
Internal connections
13 candidates
Connection 1
Tomer Kalisky
Original proposal
Proposal hypothesis: Kalisky contributes single-cell state interpretation while Shavit's announced complex-data programme motivates uncertainty-aware transfer across biological datasets.[70][127][128]
First test and score details
First test
Proposed first test: Hold out one study in a public single-cell benchmark and compare calibrated state prediction with nearest-neighbour and cell-composition baselines.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 3
Why this rank
Rank 1/13; fit 9/10 (3 topic overlap + 3 complementarity + 3 feasible first test). Preserved original co-membership proposal in o02, o04. No strictly higher-scoring candidate displaces this original. Biological direction is an announced research interest for Shavit; current laboratory operation and biological results are unverified. No automatic score boost for original membership. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Yoli Shavit biological/multimodal programme evidence is a 2025 announcement; current laboratory operation and biological experimental capability are not established. Original initiative conditions remain: o02 Yoli Shavit: Conditional later role: test uncertainty and generalization only after a specific cross-domain fidelity question is preregistered.
Connection 2
Yaara Erez
Proposal hypothesis: Erez's multimodal neural data questions fit Shavit's documented learning/uncertainty direction and experience with localization representations.[28][29][127][128]
First test and score details
First test
Proposed first test: On a public multimodal neural benchmark, hold out subjects and compare modality dropout, calibration and a linear baseline.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 3
Why this rank
Rank 2/13; fit 9/10 (3 topic overlap + 3 complementarity + 3 feasible first test). New pairing outside the frozen portfolio co-member graph. Neuroscience interpretation belongs to Erez; Shavit's announced programme does not establish a new available dataset. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Yoli Shavit biological/multimodal programme evidence is a 2025 announcement; current laboratory operation and biological experimental capability are not established.
Connection 3
Ethan Fetaya
Proposal hypothesis: Fetaya's robust geometric learning and Shavit's uncertainty/generalization direction meet directly on reliable complex-data representations.[73][127][128]
First test and score details
First test
Proposed first test: Benchmark a public multimodal model under held-out domains and missing modalities; compare calibration and robust accuracy.
Score components
- complementarity
- 2
- feasible first test
- 3
- topic overlap
- 4
Why this rank
Rank 3/13; fit 9/10 (4 topic overlap + 2 complementarity + 3 feasible first test). New pairing outside the frozen portfolio co-member graph. The two ML roles overlap; a distinct division between perturbation design and representation/calibration is needed. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Yoli Shavit biological/multimodal programme evidence is a 2025 announcement; current laboratory operation and biological experimental capability are not established.
Show 10 moreShow fewer internal connections
Connection 4
Jacob Goldberger
Proposal hypothesis: Goldberger's multimodal uncertainty methods and Shavit's representation/generalization work share a direct calibration problem across domains.[75][76][127][128]
First test and score details
First test
Proposed first test: Compare two uncertainty approaches on a public image-plus-metadata task under domain shift, keeping training data and model size fixed.
Score components
- complementarity
- 2
- feasible first test
- 3
- topic overlap
- 4
Why this rank
Rank 4/13; fit 9/10 (4 topic overlap + 2 complementarity + 3 feasible first test). New pairing outside the frozen portfolio co-member graph. Strong topical overlap comes with redundant methodological capacity, so complementarity is not maximal. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Yoli Shavit biological/multimodal programme evidence is a 2025 announcement; current laboratory operation and biological experimental capability are not established.
Connection 5
Amir Weiss
Proposal hypothesis: Amir Weiss's statistical localization and compression complement Shavit's documented transformer camera-localization work.[108][109][127][128]
First test and score details
First test
Proposed first test: On an open camera-pose benchmark, compare compressed representations with full features on pose error and calibration under scene shifts.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 3
Why this rank
Rank 5/13; fit 9/10 (3 topic overlap + 3 complementarity + 3 feasible first test). New pairing outside the frozen portfolio co-member graph. Camera and RF localization are different modalities; start with a representation-level comparison rather than assumed sensor fusion. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Yoli Shavit biological/multimodal programme evidence is a 2025 announcement; current laboratory operation and biological experimental capability are not established.
Connection 6
Orr Levy
Original proposal
Proposal hypothesis: Levy can define a biological network failure target and Shavit can propose multimodal models whose uncertainty is tested under study shifts.[5][6][127][128]
First test and score details
First test
Proposed first test: Hold out one public ageing cohort and compare an uncertainty-aware model with a sparse network baseline on calibration and stability of ranked signals.
Score components
- complementarity
- 3
- feasible first test
- 2
- topic overlap
- 3
Why this rank
Rank 6/13; fit 8/10 (3 topic overlap + 3 complementarity + 2 feasible first test). Preserved original co-membership proposal in o04. An added candidate, Yaara Erez (9/10), ranks above this original because its stated pair-specific roles and first test score higher; this original is limited as follows: Shavit's complex-biology interests are announced; avoid assuming a validated biological model or available new laboratory. No automatic score boost for original membership. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Yoli Shavit biological/multimodal programme evidence is a 2025 announcement; current laboratory operation and biological experimental capability are not established.
Connection 7
Shahar Alon
Original proposal
Proposal hypothesis: Alon can specify spatial measurement artifacts and Shavit can test multimodal representation and uncertainty under those perturbations.[60][61][127][128]
First test and score details
First test
Proposed first test: On an open spatial-expression/image dataset, hold out one section and sweep registration noise; compare calibration to expression-only inference.
Score components
- complementarity
- 3
- feasible first test
- 2
- topic overlap
- 3
Why this rank
Rank 7/13; fit 8/10 (3 topic overlap + 3 complementarity + 2 feasible first test). Preserved original co-membership proposal in o04. An added candidate, Yaara Erez (9/10), ranks above this original because its stated pair-specific roles and first test score higher; this original is limited as follows: Spatial alignment and biological expertise belong to Alon; Shavit's announced programme supports a modelling proposal, not proven tissue expertise. No automatic score boost for original membership. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Yoli Shavit biological/multimodal programme evidence is a 2025 announcement; current laboratory operation and biological experimental capability are not established.
Connection 8
Ori Ernst
Proposal hypothesis: Ernst's source-aligned text and Shavit's multimodal uncertainty direction could test when summaries should abstain if image and text evidence conflict.[93][94][95][127][128]
First test and score details
First test
Proposed first test: On a small open image-caption evidence set, inject mismatches and compare abstention and source-attribution errors with a text-only baseline.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 2
Why this rank
Rank 8/13; fit 8/10 (2 topic overlap + 3 complementarity + 3 feasible first test). New pairing outside the frozen portfolio co-member graph. This is a transfer hypothesis; Shavit's cited camera-localization work does not establish summarization expertise. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Yoli Shavit biological/multimodal programme evidence is a 2025 announcement; current laboratory operation and biological experimental capability are not established.
Connection 9
Yair Noam
Proposal hypothesis: Noam's estimation bounds could provide interpretable baselines for Shavit's camera-localization representation methods.[127][128][130][131]
First test and score details
First test
Proposed first test: On a synthetic localization problem, compare a learned estimator and a classical estimator under noise shift and report calibrated error against a stated bound.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 2
Why this rank
Rank 9/13; fit 8/10 (2 topic overlap + 3 complementarity + 3 feasible first test). New pairing outside the frozen portfolio co-member graph. Camera-pose expertise is not satellite-sensor expertise; the first test stays at an abstract estimation level. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Yoli Shavit biological/multimodal programme evidence is a 2025 announcement; current laboratory operation and biological experimental capability are not established.
Connection 10
Adam Teman
Original proposal
Proposal hypothesis: Teman's digital memory constraints provide controlled approximation settings for Shavit's proposed uncertainty-aware modelling of complex data.[1][127][128]
First test and score details
First test
Proposed first test: Quantize a small open multimodal model and compare calibration loss, prediction stability and memory footprint across held-out groups.
Score components
- complementarity
- 3
- feasible first test
- 2
- topic overlap
- 2
Why this rank
Rank 10/13; fit 7/10 (2 topic overlap + 3 complementarity + 2 feasible first test). Preserved original co-membership proposal in o02. An added candidate, Yaara Erez (9/10), ranks above this original because its stated pair-specific roles and first test score higher; this original is limited as follows: The original later-stage uncertainty role must follow a preregistered fidelity question; the announced biological programme is not demonstrated assay expertise. No automatic score boost for original membership. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Yoli Shavit biological/multimodal programme evidence is a 2025 announcement; current laboratory operation and biological experimental capability are not established. Original initiative conditions remain: o02 Yoli Shavit: Conditional later role: test uncertainty and generalization only after a specific cross-domain fidelity question is preregistered.
Connection 11
Alex Fish
Original proposal
Proposal hypothesis: Fish's memory and sensor circuits offer a hardware-error model for Shavit's generalization and multimodal learning questions.[89][90][127][128]
First test and score details
First test
Proposed first test: Perturb an open image-plus-metadata model with simulated sensor and memory errors; compare calibration and accuracy to floating-point inference.
Score components
- complementarity
- 3
- feasible first test
- 2
- topic overlap
- 2
Why this rank
Rank 11/13; fit 7/10 (2 topic overlap + 3 complementarity + 2 feasible first test). Preserved original co-membership proposal in o02. An added candidate, Yaara Erez (9/10), ranks above this original because its stated pair-specific roles and first test score higher; this original is limited as follows: The original optional modelling role needs a specific cross-domain fidelity target and access to a suitable benchmark. No automatic score boost for original membership. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Yoli Shavit biological/multimodal programme evidence is a 2025 announcement; current laboratory operation and biological experimental capability are not established. Original initiative conditions remain: o02 Alex Fish: Alternative or later circuit lead: assess memory/circuit precision and energy tradeoffs if the chosen implementation needs this expertise; security is outside the initial test. o02 Yoli Shavit: Conditional later role: test uncertainty and generalization only after a specific cross-domain fidelity question is preregistered.
Connection 12
Leonid Yavits
Original proposal
Proposal hypothesis: Yavits can expose approximation and compression choices in a genomic accelerator and Shavit can propose generalization and uncertainty checks for the output.[104][105][127][128]
First test and score details
First test
Proposed first test: On a public genomic classification benchmark, sweep approximation settings and compare held-out calibration with classification fidelity and energy proxies.
Score components
- complementarity
- 3
- feasible first test
- 2
- topic overlap
- 2
Why this rank
Rank 12/13; fit 7/10 (2 topic overlap + 3 complementarity + 2 feasible first test). Preserved original co-membership proposal in o02. An added candidate, Yaara Erez (9/10), ranks above this original because its stated pair-specific roles and first test score higher; this original is limited as follows: Shavit's biological direction is documented as an announced programme, not an established experimental laboratory; the exact fidelity question comes first. No automatic score boost for original membership. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Yoli Shavit biological/multimodal programme evidence is a 2025 announcement; current laboratory operation and biological experimental capability are not established. Original initiative conditions remain: o02 Yoli Shavit: Conditional later role: test uncertainty and generalization only after a specific cross-domain fidelity question is preregistered.
Connection 13
Hila Chalutz-Ben Gal
Proposal hypothesis: Chalutz-Ben Gal's organizational AI-adoption research could test how users respond to Shavit's uncertainty-aware multimodal outputs.[22][23][127][128]
First test and score details
First test
Proposed first test: Design paired synthetic task vignettes with conflicting modalities and compare planned measures of justified reliance before an approved study.
Score components
- complementarity
- 3
- feasible first test
- 2
- topic overlap
- 2
Why this rank
Rank 13/13; fit 7/10 (2 topic overlap + 3 complementarity + 2 feasible first test). New pairing outside the frozen portfolio co-member graph. No human-subject access or active multimodal workplace system is assumed. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Yoli Shavit biological/multimodal programme evidence is a 2025 announcement; current laboratory operation and biological experimental capability are not established.
External connections
10 candidates
Connection 1
Caroline Uhler
Massachusetts Institute of Technology
Original proposal
Uhler's official profile documents machine learning, statistics, causal inference and gene regulation, complementing Shavit's uncertainty-aware multimodal methods; this is a proposed match only.[127][128][129][497]
First test and score details
First test
On a public multimodal omics benchmark, compare predictive calibration and stability across held-out populations before evaluating any mechanistic claim.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 3
Why this rank
Rank 1/10 after semantic revision; analyst score 9 = max(1, 3+3+3): topic overlap 3/4, complementarity 3/3, feasible first test 3/3. A bounded offline comparison is specified; required datasets and domain assumptions must still be checked. Original remains first under these components; original status breaks equal-score ties only, without a prestige bonus.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified. The subject source announced a planned lab direction; current lab operation and established biological-modelling capability are not confirmed.
Connection 2
Michael Bronstein
University of Oxford
Proposed capability match: Yoli Shavit's uncertainty-aware deep learning, multimodal representation learning can be paired with Michael Bronstein's documented geometric deep learning, graph neural networks for calibrated multimodal models for genotype-to-phenotype hypothesis generation. The specific contribution is geometric structure and graph representations; this transfer is an analyst hypothesis.[74][127][128][442]
First test and score details
First test
Compare a geometry/graph-aware model with a parameter-matched unstructured baseline under structural perturbations using a public paired genotype/phenotype or multimodal biological dataset with donor and modality held out. Compare held-out error, symmetry consistency and robustness to altered graph topology with a calibrated single-modality predictor; the proposed biological lab direction remains unconfirmed operationally.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 3
Why this rank
Rank 2/10 after semantic revision; analyst score 9 = max(1, 3+3+3): topic overlap 3/4, complementarity 3/3, feasible first test 3/3. A bounded offline comparison is specified; required datasets and domain assumptions must still be checked.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified. The subject source announced a planned lab direction; current lab operation and established biological-modelling capability are not confirmed. The refreshed Oxford profile also lists an Aithyra scientific-director role; the original Oxford institution string is preserved.
Connection 3
Yarin Gal
University of Oxford
Proposed capability match: Yoli Shavit's uncertainty-aware deep learning, multimodal representation learning can be paired with Yarin Gal's documented Bayesian deep learning, uncertainty estimation for calibrated multimodal models for genotype-to-phenotype hypothesis generation. The specific contribution is uncertainty and selective prediction; this transfer is an analyst hypothesis.[78][127][128][452]
First test and score details
First test
Compare uncertainty estimates with calibrated single-model and ensemble baselines under a predefined shift using a public paired genotype/phenotype or multimodal biological dataset with donor and modality held out. Compare calibration error, risk-coverage and confident-error rate with a calibrated single-modality predictor; the proposed biological lab direction remains unconfirmed operationally.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 3
Why this rank
Rank 3/10 after semantic revision; analyst score 9 = max(1, 3+3+3): topic overlap 3/4, complementarity 3/3, feasible first test 3/3. A bounded offline comparison is specified; required datasets and domain assumptions must still be checked.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified. The subject source announced a planned lab direction; current lab operation and established biological-modelling capability are not confirmed.
Show 7 moreShow fewer external connections
Connection 4
Andrea Montanari
Stanford University
Proposed capability match: Yoli Shavit's uncertainty-aware deep learning, multimodal representation learning can be paired with Andrea Montanari's documented high-dimensional statistics, posterior sampling for calibrated multimodal models for genotype-to-phenotype hypothesis generation. The specific contribution is high-dimensional statistical baselines; this transfer is an analyst hypothesis.[127][128][479]
First test and score details
First test
Compare a regularized low-complexity estimator with a flexible model while varying sample size and dimensionality using a public paired genotype/phenotype or multimodal biological dataset with donor and modality held out. Compare generalization error, calibration and the sample-size threshold with a calibrated single-modality predictor; the proposed biological lab direction remains unconfirmed operationally.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 3
Why this rank
Rank 4/10 after semantic revision; analyst score 9 = max(1, 3+3+3): topic overlap 3/4, complementarity 3/3, feasible first test 3/3. A bounded offline comparison is specified; required datasets and domain assumptions must still be checked.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified. The subject source announced a planned lab direction; current lab operation and established biological-modelling capability are not confirmed.
Connection 5
Gregory Wornell
Massachusetts Institute of Technology
Proposed capability match: Yoli Shavit's uncertainty-aware deep learning, multimodal representation learning can be paired with Gregory Wornell's documented signal processing, statistical inference for calibrated multimodal models for genotype-to-phenotype hypothesis generation. The specific contribution is joint statistical inference and information constraints; this transfer is an analyst hypothesis.[110][127][128][503]
First test and score details
First test
Compare full-data inference with task-specific compressed statistics at fixed communication or storage budget using a public paired genotype/phenotype or multimodal biological dataset with donor and modality held out. Compare estimation error, calibration and bits per valid decision with a calibrated single-modality predictor; the proposed biological lab direction remains unconfirmed operationally.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 3
Why this rank
Rank 5/10 after semantic revision; analyst score 9 = max(1, 3+3+3): topic overlap 3/4, complementarity 3/3, feasible first test 3/3. A bounded offline comparison is specified; required datasets and domain assumptions must still be checked.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified. The subject source announced a planned lab direction; current lab operation and established biological-modelling capability are not confirmed.
Connection 6
Rahul Satija
New York Genome Center and New York University
Proposed capability match for Yoli Shavit with Rahul Satija: Single-cell integration closely fits the proposed multimodal biological direction but requires cell-level data and remains independent of whether the planned lab is operating.[127][128][492]
First test and score details
First test
Require paired cell-level modalities, donor identifiers and cell-type labels before integration. Start with a synthetic paired-cell matrix containing known types and donor shifts; compare modality-specific and integrated embeddings at identical train/test donor splits by label transfer and rare-type recovery. Do not infer cell-level structure from bulk genotype/phenotype tables or an operational laboratory.
Score components
- complementarity
- 3
- feasible first test
- 2
- topic overlap
- 3
Why this rank
Rank 6/10 after semantic revision; analyst score 8 = max(1, 3+3+2): topic overlap 3/4, complementarity 3/3, feasible first test 2/3. Single-cell integration closely fits the proposed multimodal biological direction but requires cell-level data and remains independent of whether the planned lab is operating. The experiment, numerical inputs or identity/scope needs confirmation before execution.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified. The subject source announced a planned lab direction; current lab operation and established biological-modelling capability are not confirmed. Post-review scope: Single-cell integration closely fits the proposed multimodal biological direction but requires cell-level data and remains independent of whether the planned lab is operating. This revised proposal awaits independent targeted re-review; simulated outcomes would establish model behavior only, not biological, clinical or deployed benefit.
Connection 7
Karthik Shekhar
University of California, Berkeley
Proposed capability match: Yoli Shavit's uncertainty-aware deep learning, multimodal representation learning can be paired with Karthik Shekhar's documented computational biology and genomics, neuroscience for calibrated multimodal models for genotype-to-phenotype hypothesis generation. The specific contribution is genomic and neural population modelling; this transfer is an analyst hypothesis.[127][128][436]
First test and score details
First test
Compare cell-population representations using donor-held-out rather than random-cell splits using a public paired genotype/phenotype or multimodal biological dataset with donor and modality held out. Compare population stability and held-out prediction error with a calibrated single-modality predictor; the proposed biological lab direction remains unconfirmed operationally.
Score components
- complementarity
- 3
- feasible first test
- 2
- topic overlap
- 3
Why this rank
Rank 7/10 after semantic revision; analyst score 8 = max(1, 3+3+2): topic overlap 3/4, complementarity 3/3, feasible first test 2/3. The experiment, numerical inputs or identity/scope needs confirmation before execution.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified. The subject source announced a planned lab direction; current lab operation and established biological-modelling capability are not confirmed.
Connection 8
Sarah Teichmann
University of Cambridge
Proposed capability match: Yoli Shavit's uncertainty-aware deep learning, multimodal representation learning can be paired with Sarah Teichmann's documented human cell atlases, cellular diversity for calibrated multimodal models for genotype-to-phenotype hypothesis generation. The specific contribution is cell-atlas reference and tissue heterogeneity; this transfer is an analyst hypothesis.[72][127][128][495]
First test and score details
First test
Map cell states to an independent tissue reference with donor and tissue held out using a public paired genotype/phenotype or multimodal biological dataset with donor and modality held out. Compare annotation agreement and rare-cell recovery with a calibrated single-modality predictor; the proposed biological lab direction remains unconfirmed operationally.
Score components
- complementarity
- 3
- feasible first test
- 2
- topic overlap
- 3
Why this rank
Rank 8/10 after semantic revision; analyst score 8 = max(1, 3+3+2): topic overlap 3/4, complementarity 3/3, feasible first test 2/3. The experiment, numerical inputs or identity/scope needs confirmation before execution.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified. The subject source announced a planned lab direction; current lab operation and established biological-modelling capability are not confirmed.
Connection 9
Fei Chen
Broad Institute of MIT and Harvard
Proposed capability match for Yoli Shavit with Fei Chen: Spatial biology is conditional on cell-level coordinates and an agreed biological application; a generic genotype/phenotype table supplies neither.[63][127][128][447]
First test and score details
First test
Require cell-level expression, coordinates and donor/modality metadata for a spatial branch. Until an appropriate dataset is confirmed, generate synthetic multimodal tissue with planted neighborhoods; compare spatial and coordinate-free integration using the same cell features by neighborhood recovery under coordinate jitter. The announced biological lab direction remains unconfirmed operationally.
Score components
- complementarity
- 3
- feasible first test
- 2
- topic overlap
- 2
Why this rank
Rank 9/10 after semantic revision; analyst score 7 = max(1, 2+3+2): topic overlap 2/4, complementarity 3/3, feasible first test 2/3. Spatial biology is conditional on cell-level coordinates and an agreed biological application; a generic genotype/phenotype table supplies neither. The experiment, numerical inputs or identity/scope needs confirmation before execution.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified. The subject source announced a planned lab direction; current lab operation and established biological-modelling capability are not confirmed. Post-review scope: Spatial biology is conditional on cell-level coordinates and an agreed biological application; a generic genotype/phenotype table supplies neither. This revised proposal awaits independent targeted re-review; simulated outcomes would establish model behavior only, not biological, clinical or deployed benefit.
Connection 10
Lior S. Pachter
California Institute of Technology
Proposed capability match for Yoli Shavit with Lior S. Pachter: Sequencing quantification is an upstream computational branch only when read-level inputs can be linked to the intended biological modelling task.[127][128][483]
First test and score details
First test
Audit for raw sequencing reads, barcodes, a reference and linked donor labels before quantification. Otherwise generate a small synthetic read set with known transcript counts and modality labels. Compare two quantifiers on identical reads, then the same biological representation model by count error and representation sensitivity. The biological lab remains unconfirmed operationally.
Score components
- complementarity
- 3
- feasible first test
- 2
- topic overlap
- 2
Why this rank
Rank 10/10 after semantic revision; analyst score 7 = max(1, 2+3+2): topic overlap 2/4, complementarity 3/3, feasible first test 2/3. Sequencing quantification is an upstream computational branch only when read-level inputs can be linked to the intended biological modelling task. The experiment, numerical inputs or identity/scope needs confirmation before execution.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified. The subject source announced a planned lab direction; current lab operation and established biological-modelling capability are not confirmed. Post-review scope: Sequencing quantification is an upstream computational branch only when read-level inputs can be linked to the intended biological modelling task. This revised proposal awaits independent targeted re-review; simulated outcomes would establish model behavior only, not biological, clinical or deployed benefit.
Evidence & open questions
- The planned laboratory's current operating status was not independently verified.
- No standalone CV or attributable patent record was verified.
- A genomics-specific representative publication should be added if this opportunity advances.
Inspect claim ratings and independent review
Bar-Ilan announced a Yoli Shavit laboratory intended to focus on deep learning for complex data, generalization, uncertainty and multimodal problems, and separately records her transformer camera-localization paper. [127][128]
High confidenceReview: reviewedCurrent laboratory operations, a standalone CV and a genomics-specific representative publication were not verified.
Review record
- root: supports. The July 2025 university article announces a planned 2025–26 lab focus in generalization, uncertainty and multimodal learning. It does not independently prove the lab is operational. The separate 2023 paper record confirms Shavit’s camera-localization authorship.
Calibrated multimodal genotype-to-phenotype hypothesis generation is a future research hypothesis joining Shavit's methods with Uhler's causal-genomics expertise. [127][129]
Low confidenceReview: reviewedThe genomics direction is prospective; predictive results would not establish causality, and biological validation is absent.
Review record
- root: supports. The announced multimodal/uncertainty focus and Uhler’s MIT causal-genomics profile support the proposed methods. A genotype-to-phenotype model is not an established output of this team.