Research Intern

MorphLabs · New York City · 4-month internship, on-site · Rolling applications

MorphLabs is building non-invasive telekinesis: a brain interface that turns movement intent into action, starting with people underserved by existing technology.

We are starting with continuous, intuitive prosthetic control for upper-limb amputees, decoding movement intent from EEG in real time, without surgery or reliance on muscle sensors. We develop the hardware and models together, treating signal quality, learning and control as one problem. Prosthetics are where we start, not where we stop.

We are a small, scrappy, high-throughput team in NYC. We have raised a substantial pre-seed round, with our seed round underway, and we move quickly between research and building.

The central problem at MorphLabs is learning a mapping from EEG to intended movement that holds up across people and over time. Those models drive prosthetic and robotic hands, on-screen cursors and more.

As a Research Intern, you'll spend four months on one well-scoped research project at the center of that problem. You'll work directly with our CTO and the founding team, own your project from the first question to the result, and be expected to ship something real: a decoder running in our system, a benchmark we adopt, or a clear finding that changes what we try next.

This is not a side project. Interns work on the same data, hardware and open problems as the rest of the team.

• →Transfer learning across subjects, sessions and devices. Help build decoders that adapt to new users from minimal calibration data and stay robust to drift between sessions and headsets.
• →Self-supervised pretraining. Try pretraining objectives on our growing, largely unlabeled EEG corpus, and measure how well they transfer to real control.
• →Online closed-loop decoding. Help move models from offline benchmarks to causal, low-latency decoders, and look into why offline accuracy doesn't always predict online performance.
• →Paradigm design. Help design the tasks and data-collection protocols that produce informative training data, and use each result to shape the next experiment.
• →Evaluation and benchmarking. Contribute to the metrics, held-out splits and benchmarks we use to judge decoders, offline and online.
• →Inform the hardware. Share what the models reveal about signal quality with our hardware team, and help validate our custom headset against third-party systems.
• →Ship what works. Turn promising results into reproducible decoders running in our system, and write up what worked and what didn't.

• Data, every day. We record new EEG daily, and our corpus already spans thousands of hours.
• Our own hardware. We build our own headset, electrodes and amplifier, and also work with data from third-party headsets.
• An IRB already in place, covering both healthy participants and patients.
• The resources to do the work. Compute and equipment won't be the bottleneck.

We're looking for students and early-career researchers (undergraduate, master's or PhD) who have already built things. What you've done matters more than your year, school or degree. You don't need a background in neuroscience or BCI; strong researchers can pick up the domain.

• Research you can point to, such as papers, reports or code, and a clear account of what you did yourself
• Models you built and trained yourself on noisy time series (speech, audio, sensors, robotics and neural data all count), and can explain your architecture and objective choices and how you debugged training and generalization
• Models that kept working when the data shifted, across people, sessions, devices or environments, and an understanding of why
• Rigorous, reproducible work: fair baselines, ablation studies, test sets that hold out whole sessions or people, and results others can reproduce
• Math and statistics you actually use in modeling decisions: probability, optimization, linear algebra and uncertainty
• An experiment or task you designed that produced useful data, or human-subjects research, including under IRB-approved protocols
• The ability to run experiments quickly and let each result sharpen the next one, while working independently and writing clearly

• A background in dynamical systems, geometry or information theory that shapes how you think about representations, time structure or model design
• Experience with biosignals, neural data or non-invasive BCIs
• Experience with real-time or closed-loop systems, where latency matters and offline accuracy isn't the whole story
• Experience learning from limited data: self-supervision, and choosing the next experiment when data is scarce
• Pretraining experience where the results carried over to a downstream task
• First-author papers or other substantial research results. Quality matters more than venue.

These are guidelines, not a checklist. Exceptional researchers often come from unconventional backgrounds, so apply even if you're not an exact fit.

1. Quick screen
2. Technical interview
3. Culture fit call with the full founding team
4. Chat with references
5. Offer