The Opportunity

insitro is a physical AI company dedicated to unlocking causal human biology and accelerating the delivery of better medicines to patients. Our unique Virtual Human™ platform identifies novel, high-impact genetic intervention points, which our TherML™ platform translates into therapeutics—whether small molecules, biologics, or oligos. With multiple programs in metabolic disease and neuroscience advancing toward the clinic, and our first IND submission slated for the second half of this year, we are at a pivotal inflection point.

To enable that mission, we need a software layer that ties together the scientific, automation, and machine learning platforms — that's our Lab Platform: a harness for science, built as an agentic workflow system that lets scientists drive the lab through an agent. Our bar is that everything in the lab should be agentically accessible (any action a scientist can take, an agent can take), agentically legible (agents understand what the data means in insitro's context, not just how to fetch it), and humanly verifiable (a scientist can always check what an agent did and why). This is real robotics, workflow automation, and applied agent tooling — and your users are in the building with you.

You'll work across the stack — front end, backend, and the integrations that reach into instruments, Benchling, and agent runtimes — partnering closely with our machine learning, automation, and biology teams. Based in South San Francisco, this role reports directly to Senior Manager, Software Engineering and offers an in-person hybrid schedule of three days per week. We'll bring you up to speed in the domain of drug development and back your ideas with real trust and mentorship along the way.

Responsibilities

Full-Stack Ownership

Ship Across the Stack: Build the React and TypeScript frontends scientists use every day, the Python services behind them, and integrations that reach instruments, Benchling, and agent runtimes

Own It End to End: Take a feature from the first conversation with a scientist through the data model and UI to the alert that fires when it breaks

Do the Normal SWE Things: Write code, review design docs, talk to users, and do code reviews that actually make the codebase better

Agentic Tooling & Experimentation

Chase User Appreciation: Ship things that make scientists say "OMG, this is amazing, thank you so much."

Figure Out What's Actually Useful: Test agentic tooling ideas through prototypes, real users, and short iterative loops rather than betting on theory

Move the Needle: Contribute work that meaningfully advances insitro's mission and the pace of drug development

Cross-Functional Partnership

Go Watch the Work: Spend time in the lab observing the workflow before you change it

Partner Broadly: Collaborate closely with machine learning, automation, and biology teams to build tools people actually use

Keep Humans in the Loop: Design for provenance and verifiability, so a scientist can always check what an agent did and why

About You

Experience & Qualifications

Tenure: 2–4+ years of experience as a professional software engineer

Engineering Fundamentals: Working knowledge of AWS or GCP, relational databases, and standard practices like version control and code review

Core Competencies

Product Instinct: You think like a product person — you want to know who the user is, what they're actually trying to do, and you have opinions about what to build

Curious About the Science: You don't need a biology background, but you want to learn the domain rather than treat it as someone else's problem

Comfortable in the Gray Area: You can reason clearly about the tradeoffs between quality and speed

Fits the Team: You're up for writing design docs, having opinions in code review, and yes, goofing around on Slack.

Preferred Qualifications

Domain Experience: Experience with LIMS, lab automation, or another life sciences domain

Agent/LLM Experience: Experience building with LLMs or agent frameworks — tool and skill design, evaluation, or getting a model to behave reliably against real systems