RESEARCH INTERN

Anto Bio (YC F25); San Francisco (CA); On-site; USD 140,000 - 230,000; Full time

Benefits offered by this job

Direct patient impact

World-leading researchers & advisors

Real-world scale projects

Job summary

Anto Bio is a frontier biology AI lab developing sparsification and tokenization methods to unlock microbiome data and build foundation models for microbial communities. The team includes researchers and engineers from MIT, Cambridge, Harvard Medical School, and companies like J&J and IBM Research.

We value quick learning, robust data workflows, and clear communication. The interview process includes a phone screen, a 45 minute follow-up, and a take-home task with references.

Qualifications

• Proven ability to learn new topics quickly and communicate clearly.
• Familiar with training pipelines for large models (RL, SFT, evaluation).
• Comfortable with multi-node training and related tooling.

Responsibilities

• Prototype and evaluate frontline ML models in biology AI pipeline.
• Collaborate with cross-functional teams in engineering and product.
• Document research findings and publish results.

Skills

ML research

PyTorch

Data wrangling

Documentation

Education

PhD or MS in CS/ML

Tools

TorchRun

Accelerate

CUDA

Job description

Anto is a frontier biology AI lab. We develop novel sparsification and tokenization methods to unlock unprecedented amounts of microbiome data, build foundation models for microbial communities, and translate discoveries into frontier research and clinical outcomes.

What We're Building

Anto is a frontier biology AI lab. We develop novel sparsification and tokenization methods to unlock unprecedented amounts of microbiome data, build foundation models for microbial communities, and translate discoveries into frontier research and clinical outcomes. This is an incredibly ambitious mission, but we believe that a team of ambitious people with high ownership can accomplish incredible things.

About The Team

• We’re a team of mission-driven researchers and engineers from MIT, Cambridge, Harvard Medical School and companies like J&J and IBM Research.
• We work on frontier methods that scale.
• Everyone on the team is a scientist. Everyone can code.
• We publish in leading AI x biology x microbiome venues, and are always active at leading AI conferences.

Requirements

• Think you can learn almost anything in two weeks, and occasionally prove it, without being insufferable about it.
• Familiar with modern training pipelines for large models (RL, SFT, evaluation).
• Comfortable with torchrun/accelerate/multi-node training, or confident you can get comfortable fast.
• Clever about getting the data you need – whether that’s cleaning, generating, or rethinking the problem.
• Like simple solutions to hard problems, but are willing to drop into PyTorch internals or CUDA when that fails.
• Can articulate ideas clearly. Writing docs, technical reports, and internal memos is part of the job, not an afterthought.
• Turn insights into simple, elegant user flows and polished UI that customers love.
• You should have read most of our papers (see 'Publications'), specifically:
• A. E. Gollwitzer, D. A. Subramanian, I. Tucker, and G. Traverso, “Steering the evolutionary game: Hierarchical control of therapeutic resistance in cancer treatment,” in Proc. NeurIPS 2025.
• A. E. Gollwitzer, D. A. Subramanian, I. Tucker, and G. Traverso, “MetaOmics-10T: The foundational dataset to unlock causal modeling of microbise ecosystems,” in Proc. NeurIPS 2025 AI for Science Workshop, 2025.

What we look for

• We look for people who never quite feel done. Better always seems possible.
• Startup work is intense and sometimes demoralizing. You can pour weeks into an idea and discover it doesn’t move the needle. The only failure is not learning fast enough.
• Quickly iterating to problem-solve, prototype and test new designs and solutions with Engineering, Product and small, nimble cross-functional teams.
• Driving clarity by effectively communicating design assets and building empathy for our customers among stakeholders.
• Working with scrappy teams, ship skateboards not cars.

Perks

• Your work connects directly to real patients, not just leaderboard numbers.
• Work on frontier methods. Everything we do scales in the real world, no toy benchmarks.
• We work with world-leading researchers and advisors in the space.

We run a fast interview process. After the initial phone screen, we book a slightly longer 45 minute follow up meeting within 48h. If that goes well, we extend an offer contingent on a take-home technical and references. That’s it.