Complementarity over similarity
Connections worth
testing together.
Each initiative combines distinct contributions around a shared problem. These are analyst proposals, not established collaborations or claims that an idea is new to science.
All ten concepts are discussion hypotheses. No team, data access or funding route is confirmed ready. All initiatives have low forecast confidence. Discussion order follows an explicit pilot/develop/explore stage, then a stable ID. It is not a ranking of people or grant success.
No matches. Try a name, capability or a broader search.
Trustworthy knowledge · Pilot candidate
Research AI whose evidence can be challenged
Can a research assistant preserve uncertainty and citation support when combining findings across disciplines?
The roster brings together source-aligned summarization, uncertainty estimation, robust learning, cryptographic proofs and human adoption research. NIST provides a risk-management frame; it does not certify an AI system.[22][73][75][76][93][94][95][123][157]
Capability pool
Proposed contributors; conditional roles can join after the first question is defined.
- Ori ErnstAlign each generated proposition to its exact supporting passages and expose conflicting sources.[93][94][95]
- Ethan FetayaDesign adversarial and out-of-distribution tests for misleading synthesis.[73]
- Jacob GoldbergerCalibrate uncertainty and abstention in multimodal evidence.[75][76]
- Mor WeissConditional later role: only after a protected-input threat model exists, specify a computation or access-policy statement that can be proven without revealing inputs.[123]
- Hila Chalutz-Ben GalConditional later role: design a separately approved researcher-use study of noticing and correcting unsupported recommendations.[22]
Start scoping with: Ori Ernst, Ethan Fetaya, Jacob Goldberger
External capability to add
Recruit an independent scholarly-information or reproducibility group to label evidence errors. The cryptography and NLP candidates on the member pages are possible method partners; none has agreed.
Funding routes to assess
GIF: AI in Science and Society · NSF–BSF joint research · ISF–DFG joint basic research · Industry-sponsored research through BIRAD
A German partner is required for GIF/DFG–ISF and a US partner for NSF–BSF. None is qualified for this question yet. Industry support would need a costed, independent evaluation agreement.
First experiment
Before starting: confirm the scoping team, an accountable scientific owner, access to the required data/materials and a costed protocol. Use an approved open corpus from two fields. Freeze a set of review questions, have subject experts label evidence, and compare ordinary synthesis against source-aligned synthesis with abstention. Keep private personnel data out of the pilot.
Evidence to continue
Measure unsupported-claim rate, missed counterevidence, expert correction time and calibrated abstention on a held-out set. Define acceptable tradeoffs before seeing results.
Stop or redesign if
Stop or narrow scope if evidence tracking only adds presentation without reducing substantive errors, or if reviewers cannot audit the proposed proofs.
University benefit to test. An openly documented evaluation method could make Bar-Ilan a trusted contributor to research infrastructure and improve its own planning quality.
Researcher benefit to test. A shared benchmark creates separate publishable questions in NLP, robustness, cryptography and organizational behavior, with contribution-based authorship agreed in advance.
A cryptographic proof can establish a specified computation, not the truth of a scientific claim. Novelty and practical proof overhead are unverified. This is a proposed experiment, not a deployed institutional decision system. All listed roles are proposed; feasibility readiness and funding-route readiness have not been established.
Computing × biology · Pilot candidate
Genomics within an energy budget
Can hardware-aware genomic analysis save measured energy while preserving biologically relevant conclusions?
Yavits contributes genome-oriented computer architecture; Teman and Fish contribute low-power circuit and memory methods; Kalisky and Shavit can define biological and statistical fidelity. IEA electricity scenarios motivate measurement, not a forecast that one architecture wins.[1][70][89][104][107][127][128][156]
Capability pool
Proposed contributors; conditional roles can join after the first question is defined.
- Leonid YavitsSelect an existing genomic kernel and a feasible processing-in-memory implementation.[104]
- Adam TemanMeasure digital implementation and memory-access energy under a common workload.[1]
- Alex FishAlternative or later circuit lead: assess memory/circuit precision and energy tradeoffs if the chosen implementation needs this expertise; security is outside the initial test.[89]
- Tomer KaliskyDefine cell-state or sequence-analysis fidelity against biological reference labels.[70]
- Yoli ShavitConditional later role: test uncertainty and generalization only after a specific cross-domain fidelity question is preregistered.[127][128]
Start scoping with: Leonid Yavits, Adam Teman, Tomer Kalisky
External capability to add
Onur Mutlu at ETH Zurich is a candidate for independent architecture benchmarking; the member page records the source. A separate genomics-data steward is needed for dataset and reuse permissions.
Funding routes to assess
NSF–BSF joint research · ISF–DFG joint basic research · Applied Research in Academia · Industry-sponsored research through BIRAD
The ETH candidate does not satisfy the US or German bilateral partner requirement. Identify the appropriate US/German team or prospective Israeli industrial adopter before selecting a route.
First experiment
Before starting: confirm the scoping team, an accountable scientific owner, access to the required data/materials and a costed protocol. Benchmark one public, properly licensed genomic workload on an available CPU/GPU baseline and one simulator or available accelerator. Freeze dataset versions and accuracy requirements before optimization.
Evidence to continue
Report end-to-end joules per valid result, throughput, peak memory, accuracy and calibration, including preprocessing and host-device movement.
Stop or redesign if
Stop the hardware route if savings disappear after total-system costs or if biological error exceeds the preregistered tolerance.
University benefit to test. A reusable benchmark and an independently replicated efficiency result would support credible computing-for-health visibility and industry discussions.
Researcher benefit to test. Connect architecture methods to a demanding real workload while giving biology and ML partners substantive questions about fidelity and uncertainty.
Some chip researchers already coauthor work; the proposed distinction is a new shared evaluation question. No fabrication budget, existing hardware access or clinical suitability has been verified. All listed roles are proposed; feasibility readiness and funding-route readiness have not been established.
Operations × public benefit · Pilot candidate
Hospital capacity without hidden tradeoffs
Can a hospital-flow model improve resilience while making service and workforce tradeoffs explicit?
The roster combines stochastic operations, prescriptive decision models, online allocation, network dynamics and people analytics. Ageing-related service needs provide context, while a local service owner must define the actual bottleneck.[22][64][65][79][80][97][98][99][119][120][122][155]
Capability pool
Proposed contributors; conditional roles can join after the first question is defined.
- Izack CohenBuild a queueing and process model of a narrowly defined patient-flow bottleneck.[97][98]
- Gonen SingerCompare cost-sensitive decisions under uncertain demand and capacity.[79][80]
- Ilan Reuven CohenConditional later role: add online/fair-allocation rules if the chosen service decision requires them.[119][120]
- Shahar SominConditional later role: test temporal network signals only if approved event-network data adds value beyond the queueing baseline.[64][65]
- Hila Chalutz-Ben GalEvaluate staffing feasibility and whether proposed changes shift hidden burden onto workers.[22]
Start scoping with: Izack Cohen, Gonen Singer, Hila Chalutz-Ben Gal
External capability to add
Wil van der Aalst at RWTH Aachen is a process-mining candidate and Alvin Roth at Stanford a market-design candidate. Start with a hospital operations owner who controls a suitable dataset; no hospital partnership is established here.
Funding routes to assess
NSF–BSF joint research · ISF–DFG joint basic research · Industry-sponsored research through BIRAD
RWTH Aachen and Stanford are potential German and US method partners, respectively; neither is committed or call-qualified. A hospital operations owner and approved data access come first.
First experiment
Before starting: confirm the scoping team, an accountable scientific owner, access to the required data/materials and a costed protocol. Model one discharge-to-next-service transition using synthetic data first, then an institutionally approved retrospective dataset. Compare to existing rules under normal load and a predefined demand shock.
Evidence to continue
Measure waiting-time distributions, blocked capacity, staff burden and subgroup service disparities. Report tradeoffs rather than collapsing them into one score.
Stop or redesign if
Stop if the model cannot reproduce baseline operations, if data access is unavailable, or if apparently improved averages conceal unacceptable service harm.
University benefit to test. A reproducible public-service study could demonstrate regional value and create a durable training and research partnership.
Researcher benefit to test. A common real-world decision problem supports distinct methodological contributions and doctoral projects. Existing Cohen/Cohen and Singer/Cohen links can be a starting point.
This is operational research, not a clinical recommendation. No efficiency gain, mortality effect, partner agreement or deployable hospital system is claimed. All listed roles are proposed; feasibility readiness and funding-route readiness have not been established.
Networks × cell biology · Consortium candidate
What makes an ageing tissue lose coordination?
Do changes in spatial cell relationships explain failure that is missed by cell-type averages?
Levy studies ageing and biological networks; Alon maps spatial gene expression; Kalisky analyzes single-cell states; Shavit contributes multimodal uncertainty. Their complementary methods support a basic-mechanism question before a diagnostic application.[5][60][61][70][127][128][129][155]
Capability pool
Proposed contributors; conditional roles can join after the first question is defined.
- Orr LevyFormulate falsifiable network-level hypotheses about ageing and loss of coordination.[5]
- Shahar AlonChoose a spatial sequencing design and quantify spatial measurement artifacts.[60][61]
- Tomer KaliskyDefine cell states, tissue context and biological interpretation.[70]
- Yoli ShavitCompare calibrated multimodal models against simpler cell-composition baselines.[127][128]
Start scoping with: Orr Levy, Shahar Alon, Tomer Kalisky
External capability to add
Seek an experimental ageing or perturbation lab with a compatible model system. Caroline Uhler at MIT is a candidate for causal-genomics methods, not a substitute for a wet-lab perturbation partner.
Funding routes to assess
Human Frontier Science Program · ERC Synergy · ISF–DFG joint basic research · MSCA Doctoral Networks
Identify the ageing/perturbation lab and model system. HFSP requires the international basic-biology configuration; ERC needs 2–4 complementary PIs; MSCA needs a qualified doctoral-training consortium. A German partner is still missing for DFG–ISF.
First experiment
Before starting: confirm the scoping team, an accountable scientific owner, access to the required data/materials and a costed protocol. Reanalyze one permitted spatial/single-cell dataset and nominate a small number of interactions for an independently designed perturbation. Distinguish association from causal evidence.
Evidence to continue
Require an interaction hypothesis to improve held-out prediction beyond composition and batch effects, then survive a targeted perturbation test.
Stop or redesign if
Stop the causal interpretation if effects vanish under batch correction or are explained by tissue composition alone.
University benefit to test. An openly documented cross-scale mechanism study could support a distinctive international basic-biology consortium and joint doctoral training.
Researcher benefit to test. Expand existing methods into a shared biological question while preserving each investigator’s methodological contribution.
Ageing is a broad context, not proof of a particular mechanism. HFSP fit requires a frontier basic-life-science question and a genuinely interdisciplinary international team; a tool or diagnostic proposal alone does not qualify. All listed roles are proposed; feasibility readiness and funding-route readiness have not been established.
Nanomedicine × optical sensing · Consortium candidate
Optical diagnostics that survive messy samples
Can physical sample handling and optical design improve robustness across realistic sample backgrounds?
Danielli develops magnetic-optical biosensing, Popovtzer works on functional nanoparticles, Lewi designs nanophotonics and Zalevsky develops biomedical optics. WHO AMR priorities motivate better diagnostics, but the pilot must select one assay and use case.[45][46][53][67][68][116][117][118][159]
Capability pool
Proposed contributors; conditional roles can join after the first question is defined.
- Amos DanielliChoose a validated reference assay and define background-suppression and sample-preparation requirements.[45][46]
- Rachela PopovtzerEvaluate nanoparticle functionalization and nonspecific-interaction risks.[53]
- Tomer LewiConditional later role: add a photonic component only if the baseline study identifies an optical bottleneck that it can address.[67][68]
- Zeev ZalevskyConditional later role: assess an imaging/readout change only if it targets the dominant measured bottleneck.[116][117]
Start scoping with: Amos Danielli, Rachela Popovtzer
External capability to add
A clinical microbiology or reference-diagnostics laboratory is essential. Aydogan Ozcan at UCLA is a computational-imaging candidate, subject to scientific fit and interest.
Funding routes to assess
Applied Research in Academia · Industry-sponsored research through BIRAD · NSF–BSF joint research
Choose the reference laboratory, analyte and setting first. IIA would need the current track-specific industrial-interest requirements; NSF–BSF needs a qualified US research partner. No adopter or co-funding commitment exists.
First experiment
Secure a reference laboratory, target analyte, intended setting and baseline assay first. Use approved safe reference materials to identify the dominant error or handling bottleneck. Confirm an owner and costed protocol, then compare one intervention against the baseline across predefined interferents; no patient deployment.
Evidence to continue
Measure analytical detection limits, false positives, between-run variability, sample-processing burden and per-test resource use against the reference.
Stop or redesign if
Stop if sensitivity gains require unrealistic preparation or worsen specificity; do not advance without an identified reference-lab partner.
University benefit to test. A well-validated method could support translational partnerships and useful technology transfer, with evidence stronger than an attractive prototype alone.
Researcher benefit to test. Creates a shared robustness problem spanning chemistry, nanophotonics and instrumentation, with possible translational outputs after ownership review.
Analytical performance is not clinical utility or approval. Patent records establish named records, not freedom to operate or Bar-Ilan ownership. This applied diagnostic concept is not presented as an HFSP fit. All listed roles are proposed; feasibility readiness and funding-route readiness have not been established.
Coding × hardware · Consortium candidate
Reliable edge computation under real faults
Can communication protocols and hardware fault models be designed together to preserve useful computation?
Gelles contributes interactive coding, Medina reliable hardware and online algorithms, Zehavi communications experience and Somekh-Baruch coding theory. The proposed bridge is between mathematical fault assumptions and measured system behavior.[19][57][87][100][158]
Capability pool
Proposed contributors; conditional roles can join after the first question is defined.
- Ran GellesSpecify an interactive protocol and its correctness conditions under communication noise.[57]
- Moti MedinaModel hardware faults and recovery costs in the computation and network stack.[100]
- Ephraim ZehaviChoose a realistic channel and system workload with tractable implementation constraints.[19]
- Anelia Somekh-BaruchDerive coding and reliability bounds that remain meaningful under model mismatch.[87]
Start scoping with: Ran Gelles, Moti Medina, Anelia Somekh-Baruch
External capability to add
Seek an embedded-systems laboratory for independent fault injection. A theory partner such as Michael Mitzenmacher is a candidate for algorithmic scrutiny, subject to a more specific source-supported methods review.
Funding routes to assess
NSF–BSF joint research · ISF–DFG joint basic research · Industry-sponsored research through BIRAD
No independent fault-injection lab or qualified US/German bilateral partner has been secured. Confirm the workload and member availability before selecting a call.
First experiment
Choose one civilian, non-safety-critical workload, device fault model and channel. Confirm the scoping team and current availability, then build an offline emulator with independently controlled packet noise, device faults and recovery delays. Compare a layered baseline with one joint protocol; qualify a separate fault-injection lab before hardware claims.
Evidence to continue
Measure correct completed tasks, energy and latency distributions under both assumed and mismatched fault patterns.
Stop or redesign if
Stop if guarantees depend on fault independence that the application cannot justify or if recovery overhead dominates useful work.
University benefit to test. A shared fault benchmark could support credible reliability research and industry evaluation without promising a safety-certified system.
Researcher benefit to test. Links proofs to realistic constraints and creates distinct opportunities in coding, algorithms and implementation.
Source profiles support the methods, not a proven joint architecture. Emeritus status and current capacity must be confirmed. No autonomous safety-critical deployment is proposed. All listed roles are proposed; feasibility readiness and funding-route readiness have not been established.
Wireless × resilience · Consortium candidate
Connectivity that degrades predictably
How should a sensing network adapt when spectrum, links and measurements become unreliable together?
Zaidel provides multiuser information theory, Leshem distributed learning, Noam estimation and spectrum management, and Weiss task-oriented sensing. IMT-2030 offers a future research frame, not evidence of 6G rollout or a guaranteed call.[102][103][108][111][112][130][131][158]
Capability pool
Proposed contributors; conditional roles can join after the first question is defined.
- Benjamin M. ZaidelDefine multiuser capacity and interference limits for the selected scenario.[102]
- Amir LeshemDesign distributed adaptation while testing adversarial and congested conditions.[111][112]
- Yair NoamModel estimation error and spectrum-sharing constraints, including satellite-terrestrial uncertainty where appropriate.[130][131]
- Amir WeissCompress and prioritize sensing information according to the task’s error budget.[108]
Start scoping with: Benjamin M. Zaidel, Amir Leshem, Yair Noam
External capability to add
Andrea Goldsmith is a documented wireless-systems candidate at Stony Brook. An independent network testbed or standards contributor would supply evaluation conditions; institutional interest is unknown.
Funding routes to assess
NSF–BSF joint research · ISF–DFG joint basic research · COST research networks · Industry-sponsored research through BIRAD
The Stony Brook candidate may support a US route, but interest and eligibility are unconfirmed. A German partner and a specific multinational COST networking plan are missing. COST would fund networking, not the experiments.
First experiment
Identify a civilian monitoring operator and an available simulator or testbed. Rank its hard constraints (for example latency, outage duration or localization error), fix a workload and failure scenarios, and obtain an approved costed protocol. Only then compare adaptation policies under link outages, congestion and contaminated observations.
Evidence to continue
Measure task error, outage duration, worst-case latency, energy and fairness of service under explicit failure scenarios.
Stop or redesign if
Stop if adaptation improves an average metric while violating the application’s hard latency or reliability limit.
University benefit to test. A reproducible resilience study could contribute to international network research and local continuity planning.
Researcher benefit to test. Gives communication theorists and statistical-learning researchers a shared test of when information is useful, rather than only how much is transmitted.
This may extend existing relationships; relationship novelty has not been exhaustively audited. Public-service partners, spectrum permissions and testbed capacity are not secured. All listed roles are proposed; feasibility readiness and funding-route readiness have not been established.
Biointerfaces × neuroscience · Early feasibility question
Link neural organization to measurable function
Which physical changes in an engineered neural network produce stable, interpretable functional changes?
Shefi can shape neural organization, Erez can frame systems-level function, Ozana can assess optical or acoustic measurements, and Goldberger can analyze signals. The bridge is a research hypothesis; cell models and human brain measurements are different evidence levels.[8][28][41][75][76]
Capability pool
Proposed contributors; conditional roles can join after the first question is defined.
- Orit ShefiDefine an engineered neural-culture perturbation with measurable structural consequences.[8]
- Yaara ErezSpecify a functional hypothesis and an analysis that avoids overinterpreting network activity.[28]
- Nisan OzanaConditional role: assess readout feasibility only if a defined optical/acoustic observable maps to the target functional signal in the chosen culture model.[41]
- Jacob GoldbergerConditional later role: evaluate uncertainty after a reference dataset and transparent signal-analysis baseline exist.[75][76]
Start scoping with: Orit Shefi, Yaara Erez
External capability to add
Add an electrophysiology lab able to provide an independent functional reference measurement. The external neuroengineering candidates on the member pages are discussion leads only.
Funding routes to assess
Human Frontier Science Program · ISF–DFG joint basic research · ERC Synergy
The mechanism, reference signal and international experimental partner are unconfirmed. HFSP/ ERC are thematic leads only; DFG–ISF additionally requires a German research partner.
First experiment
First the culture and electrophysiology leads must define one perturbation, target functional signal and independent electrical reference. Check whether any optical/acoustic observable is physically suitable; do not assume brain-flow sensing measures culture electrical activity. Only then confirm the scoping team, access, costed protocol and a bounded approved culture experiment.
Evidence to continue
Measure repeatability, signal specificity, agreement with the electrical reference and sensitivity to confounds.
Stop or redesign if
Stop translation if the readout cannot distinguish the intended functional change from motion, heating or structural artifacts.
University benefit to test. A carefully bounded bridge between biointerfaces and systems neuroscience could create a joint training platform and a defensible mechanistic contribution.
Researcher benefit to test. Offers a new experimental question for the platform, a measurement challenge for neurophotonics and a validation problem for systems analysis.
No regenerative efficacy or brain-interface readiness is established. Human research, clinical claims and therapeutic work would need separate scientific and institutional review; HFSP would apply only to the basic-mechanism scope. All listed roles are proposed; feasibility readiness and funding-route readiness have not been established.
Materials × water monitoring · Early feasibility question
Long-lived environmental sensor interfaces
Can an interface keep a water-monitoring sensor stable as fouling, chemistry and temperature change?
Tesler contributes surface control, Sadia functional ceramic and ionic-transport methods, and Weiss measurement under uncertainty. Global monitoring gaps motivate a question; they do not establish demand for this particular device.[12][31][108][160]
Capability pool
Proposed contributors; conditional roles can join after the first question is defined.
- Alexander TeslerDesign and characterize one antifouling surface treatment.[12]
- Yatir SadiaAssess whether a functional ceramic or ionic material can provide a chemically stable sensing element.[31]
- Amir WeissDesign estimation and calibration that expose environmental mismatch rather than hide it.[108]
Start scoping with: Alexander Tesler, Yatir Sadia, Amir Weiss
External capability to add
A water-chemistry lab and a monitoring operator must define the analyte, baseline and maintenance problem. An international surface-science partner can challenge durability claims; candidates appear on the relevant profile pages.
Funding routes to assess
ISF–DFG joint basic research · NSF–BSF joint research · Industry-sponsored research through BIRAD
Identify the water-monitoring operator and failure mode first. US and German research partners are not yet named for the bilateral routes; an industry route needs verified demand and terms.
First experiment
Before starting: confirm the scoping team, an accountable scientific owner, access to the required data/materials and a costed protocol. First select one measurable water-quality variable with an operator. Test untreated and treated sensor coupons in a controlled aging/fouling protocol; establish material-readout compatibility before building an integrated device.
Evidence to continue
Measure drift, selectivity, cleaning cycles, calibration frequency and environmental compatibility against an existing sensor approach.
Stop or redesign if
Stop if surface treatment impairs sensitivity, introduces an unacceptable material-release risk, or adds no durability benefit over a simpler maintenance change.
University benefit to test. A practical monitoring contribution could support environmental partnerships and international reproducibility work with a clear local benefit.
Researcher benefit to test. A common durability problem could connect materials transport, surfaces and statistical estimation without committing to a full device programme. Temporal optics was considered but removed because no fast-signal requirement was established.
Material compatibility, monitoring need and partner access are unverified. This is an exploratory combination; no water-treatment performance, field lifetime or product-market demand is claimed. All listed roles are proposed; feasibility readiness and funding-route readiness have not been established.
Quantum materials × photonics · Early feasibility question
From quantum materials to a usable photonic interface
Which emitter–photonic-platform pairing remains promising after realistic noise and fabrication constraints are included?
Panfil contributes colloidal emitters, Goldzak Mizrahi electronic-structure modelling, Cohen quantum-measurement theory, Blau frequency-domain control, Zektzer atom-photon interfaces and Desiatov nonlinear integrated photonics. The first decision is which pairings are physically compatible.[25][37][50][113][114][115][132][133][134][135][136][137]
Capability pool
Proposed contributors; conditional roles can join after the first question is defined.
Start scoping with: Yossef Efraim Panfil, Tamar Goldzak Mizrahi, Boris Desiatov
External capability to add
Feliciano Giustino at UT Austin, Morgan Mitchell at ICFO and Marko Loncar at Harvard are complementary candidates for materials, experimental sensing and photonics respectively. Select the capability needed after the first compatibility study; do not recruit everyone by default.
Funding routes to assess
ERC Synergy · NSF–BSF joint research · ISF–DFG joint basic research · MSCA Doctoral Networks
Downselect the emitter/platform before selecting PIs. ERC requires 2–4 PIs; MSCA requires a doctoral consortium with the eligible-country structure; DFG–ISF requires a German partner. US candidates are leads, not qualified NSF–BSF partners.
First experiment
Begin with a two- or three-person scoping team to build a compatibility matrix for at most two emitter–platform pairings. Confirm access to existing characterization data and a costed study; compare transparent noise/loss models. Add frequency-control, atom-interface or measurement-theory specialists only where the surviving pairing needs them.
Evidence to continue
Report spectral compatibility, loss, coherence/noise limits, operating temperature, fabrication tolerance and the advantage required over a classical reference.
Stop or redesign if
Stop an integration branch if its operating regimes are incompatible or any claimed quantum advantage disappears with realistic loss.
University benefit to test. A focused, independently tested interface result could anchor international quantum training and a recognizable research contribution.
Researcher benefit to test. Creates a route from separate materials, theory and device results to a shared question while preserving smaller publishable subproblems.
Six listed contributors are a capability pool, not a proposed six-PI ERC Synergy application; that route requires a focused 2–4-PI team. Platform compatibility and quantum advantage are unverified. Desiatov–Loncar is an observed existing coauthorship link, so this would extend a relationship. All listed roles are proposed; feasibility readiness and funding-route readiness have not been established.
