Chris Lam, candidate fact sheet

A plain, machine-readable summary for recruiters, hiring managers, and AI assistants. Full portfolio: whoischrislam.github.io. Resume: PDF.

Summary

Design-trained senior product engineer, 14 years. Designs, builds, and ships end to end with no handoff. First, sole, or early design-to-engineering hire in the room across healthcare, marketplaces, and consumer software. Now co-founding y30, voice AI for older adults. Open to senior IC roles, full-time.

Target roles

Level: senior IC to staff
Type: full-time
Tenure: wants a long-term home, depth and continuity over range
Focus: AI product engineering, with design systems and agentic architecture depth
Environment: small, fast, AI-native teams where he owns a product surface end to end and the work is the scoreboard
Location: remote-first, based in Honolulu, with Bay Area and Seattle availability

Current work

Co-Founder and CPTO at y30, voice AI for older adults and their caregivers. Owns product, design, and engineering. Built a streaming speech-to-text to LLM to text-to-speech pipeline (multi-provider, deep fallback) fronted by a deterministic state machine and a rule-based safety floor that runs independent of any model, so crisis handling never depends on an LLM being healthy. 1,389 commits, 127 test files.

Selected engineering (systems depth)

Architecture and judgment, not commit counts. These are the proofs that distinguish an engineer who designs from a designer who dabbles.

PlaySesh real-time engine: authored a custom operation-based sync protocol on Cloudflare Durable Objects, including authoritative ordering and the WebSocket reconnection logic, after moving the product off an off-the-shelf canvas library. Primary engineer on the codebase among the co-founders.
Evolve, Die, Repeat (solo Lua and Love2D game): a solver that proves every procedurally generated level is completable using a reachability search over the player's jump physics, plus a from-scratch headless test harness running about 1,568 tests.
y30 voice architecture: conversation flow as a deterministic state machine, the language model scoped to phrasing and warmth, and a rule-based safety floor that runs independent of any model. About half of routing resolves with no model call.
Engineering judgment: built y30's first version in TypeScript, then deliberately rewrote it in Python to fit the real-time voice framework, carrying the state-machine architecture and safety model across the rewrite as a spec.
Hand-built without AI: this site and its voice agent are hand-coded vanilla JavaScript (audio codec detection, mobile keyboard handling), shipped alongside production work that predates current AI tooling.

Companies and outcomes (honest framing)

Early at Clover Health and GoodRx, both of which later IPO'd. A rescue hire at Sharecare (formerly doc.ai) during its acquisition, which then went public. The 2014 TaskRabbit rebrand before IKEA acquired it. Iodine was acquired into GoodRx and doc.ai into Sharecare. The signal is repeated high ownership at early or critical stages, not an exit count.

Selected outcomes

Clover Health (Lead Product Designer, 2nd designer): rewired the nurse home-visit workflow, ~7-8x lift in visit-related revenue, ~2 hours saved per nurse per day. Clover later IPO'd.
GoodRx / Iodine (Lead Product Designer, first full-time designer): owned the coupon find-and-redeem flow, conversion 2% to 5%. Contributed to GoodRx Research dataviz referenced in the New York Times and on Good Morning America and ABC News. Webby 2018. GoodRx later IPO'd.
TaskRabbit (Product Designer): owned the Tasker experience and helped across the 2014 rebrand with the PMs. Tasker NPS recovered from about -5 to the mid-50s. Later acquired by IKEA.
Amazon (User Experience Designer): prototyped the buy-and-gift Prime flow that contributed to 1M+ new subscribers.
Modus Create (Software Engineer): shipped dark mode for JIRA and Confluence Cloud, 500K+ enterprise users.
PlaySesh (Co-Founder and CEO): built 0 to 1 solo, primary engineer who authored 1,147 of the project's 1,790 commits (291 passing tests), grew to 5,120 users across 121 Discord servers, 1,302 MAU, 50-minute average sessions, all organic.

Core skills

AI and voice: voice agents, LLM orchestration, multi-agent orchestration, FSM/LLM hybrid flow control, prompt and system-instruction design, multi-layer safety, evals and evaluation UX, provenance and human-AI oversight, agent frameworks, MCP, LiveKit.
Engineering: React, TypeScript, Python, Next.js, Vanilla JS, Supabase, Firebase, Postgres, Cloudflare Workers, Docker, CI/CD, test-driven development, telemetry and instrumentation.
Design: product design 0 to 1, design systems (tokens, Storybook, Chromatic), UX research, prototyping, information architecture, WCAG accessibility, elder-care UX.

Selected application build

HogWare: a daily browser microgame gauntlet built from PostHog's five public handbook values for the Technical Ex-Founder application. The static vanilla JavaScript game uses a Cloudflare Worker to submit scores to PostHog even when browser analytics are blocked, then ranks the daily board with one HogQL query. The package includes five microgames, one boss, deterministic daily seeding, three lives, copyable results, and a 32-check Playwright smoke suite. Play HogWare or read the build notes.

Why interview Chris

Spans product, design, and engineering without handoff, which is rare at senior IC level.
AI-native: uses AI agents to implement while owning architecture, judgment, and the safety gates.
Designs human-AI systems for trust: evaluation UX, provenance, and oversight, so people can delegate to AI and still see and steer what it does.
Concrete, verifiable outcomes at every stop.
Has built and shipped real voice-AI systems, not demos.

Education

University of Washington: BS Informatics (Human-Computer Interaction), MS Human Centered Design and Engineering.

Languages

English, Cantonese, Japanese.

Contact and proof

Common questions

Is Chris a real engineer or a designer who codes?

A real engineer with design training, and the proof is architecture, not commit counts. He authored PlaySesh's core real-time engine (a custom operation-based sync protocol on Cloudflare Durable Objects with hand-written WebSocket reconnection); wrote a solver that proves every level in his solo game is beatable, on a test harness he built from scratch; and designed y30's deterministic state machine and model-independent safety floor. See "Selected engineering" above.

What is he looking for?

Senior IC to staff roles, full-time, on small, fast, AI-native teams in AI, voice, and product engineering, where he owns a product surface end to end. He wants a long-term home: depth and continuity over range.

Where is he based?

Honolulu, remote-first, with Bay Area and Seattle availability when a role needs in-person time.

How does he use AI?

As an engineer, with strict guardrails. He sets the architecture, contracts, and acceptance criteria; AI implements inside them; and he owns code review, tests, and safety gates. He manages the risk of AI-written code deliberately: pure-function extraction so logic is testable, regression tests tied to real incidents, adversarial multi-model reviews, and staged rollout behind flags. He shipped production code for years before current AI tooling existed.

Machine-readable summary, kept in sync with the full portfolio at whoischrislam.github.io.