Founding Engineer (In-Person)

About Raycaster

Raycaster is building AI agents for life sciences - starting with the most painful part of drug development: regulated documentation. We help teams keep protocols, Module 3, CMC, quality, and labeling documents in sync with the science so submissions move faster and with fewer surprises.

Backed by YC and leading life sciences investors, we’re already working with partners like Agilent, Yokogawa Life Sciences, and other top biopharma and tools companies. Our stack is the infrastructure layer for AI-powered document workflows in biotech - think Microsoft Word × Git × Claude Code, for regulated work.

If you want a deeper dive on what we’re building and why, Sacra did a full writeup here:
Sacra research on Raycaster - https://sacra.com/research/levi-lian-raycaster-vertical-ai-workflows/

We’re starting in life sciences, but the core problems we’re solving - AI for high-stakes, document-heavy, regulated work - are much broader. Over time, we may expand into other domains (financial services, legal, safety-critical industries), so we’re looking for someone who can pick up new domains quickly and cares about product quality in any vertical.

The Role

As a Founding Engineer, you’ll own core product and infrastructure in a space where “good enough” isn’t good enough. This is a zero-to-one role: you’ll shape the architecture, ship production systems, and help define what “AI for regulated documents” even means as a category.

You’ll work directly with the founders on:

• AI Agent Architecture
  Design and optimize LLM-based agents: tool orchestration, retrieval, safety/guardrails, and sandboxed execution environments that operate on real customer documents.
• Systems & Infrastructure
  Build distributed systems for version control, impact mapping, real-time collaboration, and document-processing pipelines that must be correct, observable, and resilient.
• Full-Stack Product Work (with real product taste)
  Ship features across our Go backend and React frontend: schema design, APIs, performance work, and the kind of “it just feels right” UI/UX that expert users rely on all day.
  We care a lot about design and interaction details: copy, states, spacing, how something feels under a mouse. You don’t have to be a designer, but you should have opinions.
• Research → Production (with room for actual research!)
  Take ideas from papers / prototypes (RAG, embeddings, agent reasoning, document understanding) and turn them into reliable production workflows.
  When it makes sense, we also want to:
  • Design new evaluation methods for agent reliability in regulated workflows (working on a paper right now)
  • Build and open up benchmarks / datasets for document-heavy, high-stakes tasks
  • Publish technical deep dives, benchmarks, or papers that become reference points for the industry

This role is for someone who wants to own problems end-to-end and is happy when the line between “research” and “product” is blurry.

What You’ll Do

• Architect and implement core services for agents, workflow orchestration, and document versioning
• Design data models for complex regulated content (dossiers, protocols, specs, variations)
• Build interfaces that let scientists and regulatory teams trust and control AI outputs
• Define how we evaluate agents (beyond toy benchmarks) and ship those evals
• Collaborate on technical deep dives, benchmarks, and (occasionally) papers or talks
• Partner with customers to understand workflows and translate them into product and APIs
• Help with early hiring and set the technical + research + product bar for the team

You Might Be a Fit If You Have

• Strong systems engineering background
  • Significant experience in Go, Rust, or a similar systems language
  • Comfort with distributed systems, concurrency, and performance tuning
• Hands-on experience with production AI/ML systems
  • Deployed LLM, RAG, or model-driven systems to real users
  • Thoughtful about evaluation, guardrails, and failure modes
• Researchy instincts, builder bias
  • You like reading papers, but you like breaking and reassembling them into products even more
  • You’re excited by questions like: “what is the right benchmark for regulated-agent reliability?”
  • You’re happy to publish or open-source when it’s genuinely useful, not just for badges
• Product + design sensibility
  • You notice when a layout is off by 4px or the copy doesn’t match the mental model
  • You can pair with a designer and also make reasonable product/design calls yourself
• Curiosity about complex domains
  • Interested in how big, messy organizations actually work beneath the org chart
  • Enjoy modeling gnarly workflows into clean abstractions

Signs This Might Be You

• If you don’t know how to do something, your default reaction is:
  “Give me a few hours, a spec, and a keyboard.” And then you actually figure it out.
• You have an unsatiable appetite for learning - new domains, tools, OSS projects, weird papers.
• You have side projects, repos, or writing where you went way deeper than anyone asked you to.
• You care about the whole product: docs, error messages, onboarding, not just the clever algorithm.

Nice to Have

• Experience with LLM agents, RAG systems, or document AI
• Background in version control, CRDTs, or collaborative editing systems
• Familiarity with biotech/pharma (regulatory, QC, CMC, clinical, safety, or labeling)
• Open-source work, technical blog posts, or papers / talks that show how you think

Why This Is Interesting

• Hard, important problems.
  You’ll work on agent reliability, document understanding, impact mapping, and real-time collaboration - for workflows where correctness actually matters and “hallucinations” are not a cute demo, they’re a risk.
• Real research surface area.
  This isn’t a pure research lab, but there’s a lot of greenfield:
  • What does a “good” AI agent for regulated documents even mean?
  • How do you benchmark that in a way serious customers (and eventually regulators) would take seriously?
  • How do you design evals and datasets that aren’t toys?
    If you want to co-create some of those answers - and occasionally turn them into posts, benchmarks, or papers - this is a good place.
• Edgy in the right way.
  We’re not trying to build another “AI note-taking app.” We’re going after the boring, high-friction parts of critical industries that nobody wants to touch because they’re hard and regulated.
  If you’re optimizing for interesting problems, strange constraints, and real users, you’ll probably have fun.
• Founding-level ownership.
  You’ll meaningfully shape the product surface area, technical architecture, and early engineering culture. This is the kind of role where, a year from now, you can point at whole subsystems and say, “yeah, that exists because I was there early.”

Location & Working Style

• NYC-based
• In-person, 5 days per week (we often work on the weekends, too 😊 )

Compensation

• Competitive base salary
• 2–10% equity (we want you to feel like an owner)
• Standard benefits

Interview Process

• Technical screen (45 min)
  Systems design and coding discussion focused on how you structure and ship real systems.
• Deep dive (60 min)
  Walk us through a complex system you’ve built. We’ll dig into architecture, tradeoffs, failure modes, and what you’d change with hindsight.
• Pairing session (90 min)
  Work together on a real-ish problem from our codebase or a close analogue - expect a mix of reading, extending, and refactoring.
• Founder chat
  Culture, values, and your goals. We’ll talk about what you want next in your career and whether this aligns.

If you’re excited about building the AI infrastructure that keeps high-stakes work in sync with the science (and everything that comes after it) - and you want a role where research, product, and weird edge cases all collide - we’d love to talk.