Nuance Research Fellow

Nuance Labs is building photorealistic, real-time AI avatars with emotional intelligence: a full-duplex audiovisual system that can listen, speak, react, interrupt, and respond like a real person.

We're a Series A company ($60M raised) backed by Lightspeed, Accel, South Park Commons, NVentures, and Define Ventures, with PhDs from MIT, UW, Oxford, CMU, and Johns Hopkins, and industry experience from Apple, Meta, Amazon AGI, and Discord. The team is small, the work is real, and the problems are unsolved.

Most conversational AI avatars today are hacks — a face slapped on a speech-to-speech pipeline, stuck in the uncanny valley: emotionless, mechanical, one-turn-at-a-time. Current systems take 2–5 seconds to respond; natural conversation requires sub-500ms. That's a 10x improvement, and it demands rethinking the entire stack.

That rethinking starts with full-duplex: an AI that listens and speaks simultaneously, perceives emotion in real time, and responds with a face that actually reflects it. It's an extremely hard problem, and we're developing foundation models designed for it from the ground up.

The Nuance Research Fellowship is a 3-month engagement for early-career researchers who want to do serious work at the frontier of multimodal AI — and to find out, alongside us, whether Nuance is where they want to spend the next chapter of their career.

You’ll contribute directly to one or more of our research workstreams: pretraining, post-training, RL, evaluation, data, multimodal modeling, or inference, working alongside people who’ve built these systems before. At the end of three months we’ll decide together whether to convert to a full-time Member of Technical Staff role, with enough lead time for you to plan your next step either way. Fellows who convert step into MTS-level scope and ownership from day one.

We build photorealistic digital humans with full-duplex, real-time audiovisual interaction. There are open problems across the entire ML stack — training omni models from scratch, aligning them, evaluating real-time conversational behavior, shipping at sub-500ms latency. Plenty are still unsolved, and we want people who find that exciting.

• →Contribute directly to a research workstream from week one — pretraining, post-training, RL, evaluation, data, multimodal modeling, or inference
• →Read papers, reproduce key results, and bring promising ideas into our stack
• →Run experiments rigorously: design, instrument, debug, and draw conclusions from training and eval runs
• →Build evaluation harnesses, benchmarks, and analysis tooling the team relies on
• →Take research-grade prototypes and turn them into something that ships
• →Work closely with senior researchers and engineers across the team; ramp on the stack fast
• →At the end of three months, decide together whether full-time MTS is the right next step

• Strong working knowledge of PyTorch and deep learning — you can train a model, debug a training run, and reason about what’s happening at the loss level
• At least one first-author paper at a tier 1 venue (main conference proceedings) — NeurIPS, ICML, ICLR, CVPR, ICCV, ECCV, ACL, EMNLP, NAACL, ICASSP, Interspeech, MLSys, SIGGRAPH, or equivalent
• Genuine interest in joining Nuance full-time after the fellowship. This isn’t an internship or a side project — we are looking for long-term partners on this journey

• BS, MS, or PhD in CS, ML, math, physics, EE, or a related field — completed or in the final stretch
• Strong programming ability and software engineering instincts
• High agency — when you see something broken or slow, you fix it; when you see an opportunity, you take it before being asked
• A bias toward shipping over polishing, with the judgment to know when each matters
• The appetite to pick up anything and optimize the hell out of it

• Multiple tier 1 publications, or a paper that received significant attention (best paper award, broad adoption, high citation impact for its age)
• Olympiad medals or finalist-level results in IMO, IPhO, IOI, IChO, IBO, IMC, or equivalent
• Codeforces grandmaster, ICPC world finals, Putnam fellow, Kaggle grandmaster, or similar
• Hands-on experience with multimodal models (audio, video, language) or real-time systems
• Open-source contributions to major ML frameworks or research codebases
• A track record of independent projects that made something noticeably faster, smaller, or better

• Location: In-person in Seattle, five days a week — we believe in the compounding value of working shoulder-to-shoulder.
• Visa sponsorship: We sponsor visas (O-1, H-1B, green card) from day one.
• AI-native tooling: Do your best work with the best tools, including unlimited tokens.