Founding AI Engineer (YC W24) w/ .25% - 75% Equity
Bounty Amount: $9,000-$13,500
Company Name: Sonia
Role Type: Full-time
Location: San Francisco, CA (in-person)
Salary / Hourly Rate: $120,000 - $180,000 per year
Benefits: In-person SF team; high ownership on a 5-person, seed-stage startup.,Meaningful equity; fast path to scope as the product/function owner.,US Citizens Only,Open to Exceptional New Grads
Role Information
Role Overview: N/A
Responsibilities: Ship end-to-end LLM features for voice + text therapy sessions: prompt/agent design, tool use/function calling, latency & turn-taking handling, and production deployment., Build offline and online eval loops (unit tests, regression suites, shadow/prod checks) that track reliability and user outcomes (e.g., anxiety-score trends like GAD-7), and use them to decide what ships., Implement and iterate safety systems: risky-response detection, fallback/deferral strategies, human-in-the-loop escalation, and post-incident reviews; treat safety as a first-class product feature., Own Python services and lightweight product surfaces (internal tools, small UX hooks) that speed up experimentation and founder feedback loops., Partner with Design (voice/chat UX) and Clinical Research to translate findings into product improvements and safeguards; instrument what you ship so we can learn quickly in production., Drive a weekly shipping cadence: prototype → evaluate → harden → release; document decisions and metrics so the team can build on them.
Qualifications: Strong Python plus hands-on prompt/LLM engineering (tool use, function calling, retrieval or memory patterns, evals); you’ve shipped something real users touched., Product sense and speed: you can simplify ambiguous problems, choose pragmatic baselines, and deliver value in days—not quarters., Track record of safety-critical thinking (red-flag detection, guardrails, fallback paths) and comfort being accountable for quality in production., Evidence-driven mindset: you instrument features and are comfortable tying quality to measurable outcomes (not just engagement)., Collaboration in a tiny, high-trust team: you like building in person with founders and cross-functional partners (design/research). In-person, San Francisco required; w/ US work authorization (with no sponsorship needs)
Minimum Requirements: In-person, San Francisco (team works on-site). (From founder email + YC page),US work authorization (YC lists “US citizen/visa only”).,Strong Python & Swift Mobile Development,Evidence of shipped LLM/agent features (code or live demo).,Safety + eval mindset (guardrails, pre-delivery checks) given the mental-health context
Screening Questions: Please share github:,2–3 minute video (Loom or Google Drive): Record a 2-3 minute video introducing yourself and screen-sharing one project you’re proud of. If possible, keep your camera on so we can see how you communicate.,Why Sonia? What about our mission (building a safe AI therapist, voice + text) and our in-person SF culture resonates with you?
Company Information
About Company: N/A
Culture: N/A
Additional Information
Interview Process: Step 1: Vibe check call — 15 minsQuick intro, alignment on the mission and role, confirm in-person SF and work-auth fit.Step 2: Technical take-home — ~4 hours (compensated)Build a small LLM feature or eval pipeline relevant to voice + text therapy (prompt/agent design, tool use/function calling, safety checks). Include notes on trade-offs and what you’d test next. Step 3: Technical deep-dive & code review — 30–45 minsWalk through your take-home and 1–2 shipped LLM projects. We’ll dig into prompts, evals, guardrails, latency/turn-taking handling, and how you measured quality.Step 4: In-person trial in SF — 1–2 days (paid)Work on a scoped feature with the founders: prototype → evaluate → harden → ship. We look for clarity, speed, product sense, and how you reason about safety and measurable outcomes. (Team works on-site; trials are in San Francisco.)Step 5: OfferFast debrief and references as needed; discuss compensation/equity within YC-listed ranges for this role.
Day to day: Design, build, and ship end-to-end LLM features that power Sonia’s voice + text therapy sessions. You’ll iterate on prompts/agents, wire in tool use/function calling, stand up offline/online eval loops, and harden safety guardrails before releasing to production. Expect a fast idea → prototype → evaluate → ship cadence with founders, and instrumentation that ties quality to real outcomes (e.g., reliability and early GAD-7 movement the team tracks).
Team: Work directly with the three founders on a ~5-person team in person in San Francisco. You’ll pair with design on voice/chat UX and with research on measurement and risk mitigation, keeping feedback loops tight and decisions pragmatic. (YC lists skills as Prompt Engineering, Python; the culture emphasizes on-site collaboration.)
Growth: As one of the first engineers, you’ll define technical standards for LLM quality, evals, and safety, influence architecture choices, and build internal tooling that speeds experimentation. As the product and team scale, there’s scope to lead major lines (e.g., new session modalities, memory/eval systems), mentor future hires, and help formalize practices that keep Sonia outcome-first and safe—consistent with the company’s “develop it like a drug” mindset.
Ideal Candidate Profile: We’re looking for a design–engineering hybrid who owns discovery → UX/UI → implementation for our iOS app, iterates quickly with the founders in person in San Francisco, and cares deeply about measurable outcomes and safety.What makes you a strong fit • You’ve shipped mobile product end-to-end (portfolio shows discovery, design, and hands-on build), ideally in Swift/SwiftUI. • You can design stateful conversational experiences (voice + chat) and instrument what you ship to learn quickly. • You use research and data to decide—e.g., you can explain how you’d evaluate changes with validated measures like GAD-7, not just engagement metrics. • You thrive on small-team pace and high ownership, collaborating daily with founders in person in SF. • You treat safety as a product feature and can describe guardrails you’ve designed for sensitive contexts. Benchmark profile (from founders) • Design: young, hungry, ideally with some engineering skills or interest; new-grad OK. • https://x.com/floguo (design engineer; strong product taste + build skills).Signals we’re likely to pass • Only visual polish with no shipped, user-validated work. • Can’t work in person in San Francisco. • No experience designing or building for voice/chat or other safety-critical UX.Why this is exciting • Mission with evidence: Sonia is building a safe AI therapist (voice + text) and publicly emphasizes outcomes; you’ll shape how the app looks, feels, and measures progress from day one.
