Software Engineer

HorizonAI Talent; San Francisco (CA); On-site; USD 150,000 - 250,000; Full time

Job summary

HorizonAI Talent in the San Francisco Bay Area is seeking a hands-on senior full-stack SWE with 3–7 years building AI/agentic products in production. You will own from model layer to user interface for physician-facing workflows, requiring strong CS fundamentals and a startup mindset.

Responsibilities include building agentic workflows, designing eval frameworks, and improving data ingestion and context assembly for agents. Visa transfers and relocation may be considered for the right profile.

Qualifications

• 3–7 years of full-stack SWE experience with AI/agentic products in production.
• Strong CS fundamentals; bootcamp-only won’t suffice.
• Early-career time at a big tech company followed by startup experience.
• Willing to be IC builder rather than moving into management.

Responsibilities

• Build and ship new agentic workflows automating complex casework for physicians and ops teams.
• Design eval frameworks to validate agent outputs before shipping.
• Improve how platform ingests, indexes, and assembles case context for agents.
• Develop tooling and UIs to integrate agent capabilities into daily workflows.
• Set patterns and tooling for AI-native development used by the team.

Skills

Full-stack SWE

AI/agentic systems

Production experience

Frontend development

Backend development

Education

CS degree or equivalent

Job description

San Francisco Bay Area · 3 days onsite (Tue-Thu) · $150K-$250K base + equity (flex to ~$270K for the right profile) · Visa transfers considered · Relocation flex for senior candidates

The short version

I'm working exclusively with a Series A company at the intersection of medicine and law. They've done $10M+ ARR in under 14 months, are profitably growing and on track for $15M+ ARR this year - backed by top-tier healthcare investors, currently ~30 people and looking to add 3 engineers by the end of the year, with headroom for up to ten more if the trajectory holds.

The founding team is genuinely cross-domain: 30+ years of running independent medical practices on the clinical side and product/engineering leaders out of Meta, Instagram, Dropbox, Coda and Nomad who've previously scaled companies to $1B+ valuations. They're deliberately staying stealth until later this year, then coming up for air.

This is a hands-on IC seat building agentic workflows in production - the kind of role where you own the problem from the model layer through to the UI a physician actually clicks.

What they're building

They're automating the medical-legal edge of healthcare - starting with physicians managing workers' compensation claims, where the paperwork and back-and-forth with insurers is a significant tax on the time physicians can actually spend on care. The wider mission is bigger: helping physicians run and grow their own independent practices, on the view that the future of healthcare is driven by individual physicians rather than large bureaucratic systems. Thirty-year horizon.

Distribution today is 100% referral. No marketing. The pain point they're solving is real enough that clinicians are pulling the product through the door themselves.

What you'd actually do

• Build and ship new agentic workflows that automate complex casework for physicians and ops teams.
• Design and iterate on eval frameworks - the systems that decide whether an agent's output is reliable enough to ship. This is treated as first-class engineering work, not an afterthought.
• Improve how the platform ingests, indexes, and assembles case context for agents to work from.
• Build the tooling and interfaces that bring agent capabilities into users' day-to-day workflows.
• Set patterns and tooling for AI-native development that the rest of the team builds on.

Stack is deliberately language-agnostic; what matters is that it's modern. If your last five years have been in a legacy enterprise or defence-contractor stack, this won't be the right shape.

Who they're looking for

3-7 years of full-stack SWE experience, with genuine time spent building AI/agentic products in production. Above seven, they can flex for the right shape, but the seat is squarely aimed at engineers still in IC-builder mode rather than heading into management.

The career shape that clears cleanest:

• Strong CS school (adjacent - maths, engineering - is fine; bootcamp-only won't clear).
• Early-career time at a big-tech company (Meta, Airbnb, Google, etc.).
• Then a move into a startup - this is the load-bearing part. Big-tech-only careers, or big-tech-into-more-big-tech, likely won't clear.

You should be full-stack in the honest sense - you don't need to ship 50/50 backend and frontend today, but you need to be willing to reach into the 20% you're less strong in. Backend-only-and-unwilling doesn't work; frontend-only doesn't either.

On the AI side, they mean "in production." Hands-on with frontier LLMs, and - this is the sharpest gate - eval frameworks and guardrails running in production. LLM-as-judge, golden datasets, regression gates, guardrails on live agent output. Not "used ChatGPT once." Not "familiar with evals from the Anthropic cookbook." Shipped.

Other things that matter:

• You take vague problems to shipped solutions without needing a spec or constant direction.
• You engineer for correctness and reliability, not just functionality - this is a domain where an unreliable agent is worse than no agent.
• You genuinely want the pace and ambiguity of an early-stage startup. The HM's own words: "it's extremely rewarding and challenging and fun, but it is hard."

Bonus - not required, but a real thumb on the scale: prior time in healthcare, legaltech, insurance, or another heavily regulated industry.

What "great" looks like

There's a distinction here worth knowing up front, because it's the lever on the pay band:

• Good - executes independently, owns a vertical, learns the domain on the job. Lands in the standard band.
• Great - brings a creative framework to problems, has a strong network and has directly relevant AI product and infrastructure experience. This is the profile that unlocks the top-of-band comp and, where relevant, flexibility on relocation.

Worth being honest with yourself about which side you land on when we talk.

What won't work

• Bootcamp-only origin.
• Senior engineers who've moved into management - they want IC builders, not people who've stepped away from the keyboard.
• Junior + big-tech-only - they've tried this shape and it hasn't worked. Junior + startup is fine.
• 10+ years at a single big-tech (any of the FAANG-shape names). Mode mismatch.
• Serial short stints without a story. Two short stints in a row where "both companies lost funding" is fine - that's the story. Three-in-a-row with no explanation is the problem.
• Career histories in traditional banking, defence contracting or slow-moving enterprise/large consulting.
• Currently at an obviously top-tier AI-native shop and leaving after less than a year - expect to be asked why. Not disqualifying, just needs a real answer.

The details

• Location. San Francisco Bay Area, 3 days onsite Tue-Thu. Currently in a San Mateo WeWork; own SF office by end of year. Most of the team commutes in from SF. Default is local; for the right senior profile with directly relevant background they will consider relocation - ask me and I'll get you a straight answer per candidate.
• Comp. $150K-$250K base + equity. They'll flex up to ~$270K for candidates who hit the "great" bar. Verbatim from the HM: "we're not trying to compete with Anthropic." Read: strong Series A comp, not frontier-lab comp.
• Visa sponsorship. Open, candidate-dependent. Not a blanket policy but not a hard exclusion either.
• Team size. Engineering team of 5 today (including a hands-on CTO co-founder). Hiring three to reach eight, with the intention of running a lean core team long-term.

Why this one is genuinely worth a conversation

• Profitable, growing, not raising in a hurry. Their most recent round was >$2M and, in the HM's words, "we're in no rush to raise a Series B, we feel like we're in a really good spot." That's a different pressure profile from most Series As.
• Founding team you'd actually want to work under. Prior startups at $1B+ valuations, real IPO experience and engineering out of Dropbox/Instagram/Meta paired with 30+ years of clinical operator depth. Not the usual half-team.
• A real domain, not a demo. Workers' comp is unglamorous, which is exactly why the wedge works - physicians will pay for anyone who can take the paperwork off their plate. Ten million ARR in fourteen months, on referral alone, isn't accidental.
• Foundational hire. They're going from 5 to 15 engineers, then holding lean for a long time. The patterns and tooling you set on AI-native development are what everyone builds on afterwards. That's a genuinely different ownership surface from being hire #40.
• The AI work is the product. Evals, guardrails, agent orchestration - not a bolt-on. If you've been the person quietly building the eval harness that lets your team ship agents at all, this is the seat where that work is the job, not the tax.