Applications for the 2026-2027 cohort
Transfyr AI Fellowship
A 1 year research fellowship for exceptional graduate students to lead a project solving frontier machine learning problems that impact real-world scientific systems.
Why this exists
Science is still missing an observability layer.
Scientific work is rich with tacit decisions, failed attempts, operator judgment, noisy instruments, and context that never makes it into a protocol. The next generation of scientific AI needs to reason over that reality, not just polished papers and clean benchmarks.
We're building the world's largest multimodal dataset of scientific execution - we're inviting you to come play with it.
Research agenda
Frontier problems with real scientific stakes.
Fellows will work on ML systems that need to understand scientific work as it actually happens: across people, instruments, protocols, repeated attempts, automation, and real outcomes.
Multimodal reasoning
Build models that align video, audio, protocol text, instrument traces, timestamps, gaze, and outcomes into a coherent account of what happened.
- Cross-modal grounding
- Temporal event extraction
- Human activity understanding
Conflicting evidence
Real systems disagree. Fellows will explore how models represent, weigh, and explain conflicting signals across people, sensors, protocols, and results.
- Uncertainty and calibration
- Evidence attribution
- Contradiction-aware reasoning
Long-context scientific memory
Lab work unfolds over hours, days, and repeated attempts. The research challenge is deciding what to remember, compress, compare, and retrieve.
- Long-horizon state tracking
- Retrieval over process histories
- Failure and deviation memory
Real-world evaluations
Design evaluations where success means better transfer, fewer hidden errors, stronger operator training, better automated execution, and more reproducible scientific outcomes.
- Benchmark design
- Process-level metrics
- Lab-grounded model assessment
Tacit expertise
Model the invisible parts of expert work: attention, sequence, heuristics, small corrections, and the judgment calls that separate written protocol from reproducible execution.
- Expert-novice comparison
- Skill representation
- Human-in-the-loop feedback
Physical-world AI
Help bridge foundation models, robotics, and lab work by studying how models and agents can learn from demonstrations in environments that are variable, constrained, and consequential.
- Embodied data pipelines
- Transfer across settings
- Policy and assistance evaluation
Biosecurity and bio-risk observability
Static checklists cannot secure frontier biology. Fellows will develop ML systems that monitor live laboratory workflows to evaluate real-world capability, design dynamic evals, and autonomously identify risks that can be mitigated through technical guardrails and safety-critical containment architecture.
- Synthesis-to-execution verification
- Emergent capability tracking
- Autonomous vulnerability detection and patching
Scientific execution ontology
Disentangling biological signal from procedural artifacts requires more than labels. Fellows will build ontology-driven systems that capture protocol metadata at the level experimentalists and automation experts actually reason: stock prep, concentrations, solubility limits, freeze-thaw history, plate setup, curve shapes, and execution context, so models can surface real phenotypes instead of trusting noisy growth/no-growth labels.
- Wet lab-grounded metadata design
- Cross-domain causal attribution
- Multi-scale variability mapping
The fellowship
A research home for ambitious graduate students.
Fellows join full-time for 12 months to lead a focused research project, contribute to the real solution, and work closely with scientists, operators, and ML researchers building tools that make scientific execution observable, reproducible, and automatable.
What you will do
Define a research question, build prototypes, analyze real multimodal scientific-execution data, pressure-test evaluation ideas, and produce a publishable or demo-grade research artifact.
Mentorship and team
Fellows receive direct mentorship from Transfyr's founders, technical team, and advisor network across AI, biology, robotics, scientific operations, and applied R&D.
What you bring
Strong ML fundamentals, evidence of exceptional research taste, comfort with ambiguity, and curiosity about how science actually happens in physical environments.
Who should apply
Current graduate students able to take leave, recent PhDs, and postdocs in machine learning, computer science, computational biology, robotics, human-computer interaction, statistics, or adjacent fields.
Program terms
Frontier AI for Frontier Science
This is a full-time fellowship, not a side project. Fellows get a competitive package, close mentorship, and room to build work that can become public research.
Compensation
Fellows receive $125,000 annual compensation plus benefits.
Time commitment
Fellows are full-time for 12 months. Current students should be able to self-certify that they can take leave or otherwise commit full-time.
Location
The fellowship is designed around full-time in-person work in the Boston/Cambridge area. Remote arrangements may be considered by exception when the project and candidate make it workable.
Visa support
International applicants are welcome. Visa support is available for selected fellows.
Research output
Fellows should aim to produce a publishable research artifact, benchmark, dataset, demo, prototype, or equivalent technical contribution, and will have the opportunity to implement promising solutions into real-world contexts and products. Transfyr owns fellowship work product, and supports publication subject to confidentiality and IP review.
Compute and research support
Fellows will have access to compute, GPUs, data, and project support. Tell us what resources your research proposal needs.
Dates
Application timeline
Applications are reviewed on a rolling basis. Apply by August 15 for full consideration, and include links that make your work easy to evaluate.
Rolling review
Applications reviewed as they arrive
Full consideration
Apply by August 15, 2026
Decisions
Rolling through late August 2026
Fellowship begins
Around September 2026
Apply
We can't wait to get to know you.
Application inbox
ai_fellowship@transfyr.ai
Application materials
Efficient 1-page application process. Tell us what you want to work on and what resources you need.
- Current graduate students, recent PhDs, and postdocs can apply
- Applicants should be available to start around September 2026
- Selected fellows are expected to be based in Boston/Cambridge full-time
- Include the compute, GPU, data, or support your project would need
- References will be requested of finalists