Senior Product Engineer, AI Assistant Platform (SF)

TechTree; San Francisco (CA); Hybrid; USD 180,000 - 240,000; Full time

Benefits offered by this job

Equity

Health insurance

Dental

401(k)

PTO

Gym budget

Lunch

Job summary

TechTree in San Francisco builds an AI assistant that connects to tools like Gmail, Calendar, Slack, and Notion, running 24/7 in the cloud. You’ll own the frontend surface of the assistant, ship end-to-end features, and iterate from feedback to improve user experience across surfaces.

The role emphasizes React, TypeScript, and a stack including TanStack, Zustand, TailwindCSS, IndexedDB, and Cloudflare tech.

Qualifications

• Shipped complex web products with ownership after launch.
• Experience with multiplayer or large-data apps with real fetching and caching.
• Strong React expertise and reusable component design.
• Comfort owning full-stack feature end-to-end (migration, route, fetch, UI, analytics).
• Interest in performance, caching, invalidation, and optimistic updates.

Responsibilities

• Own the product surface of the assistant: interfaces users use to hand work to it and monitor results.
• Turn rough ideas into working features, ship first version, iterate from feedback.
• Collaborate across frontend, API, and analytics to ship end-to-end features.
• Build routines, apps, and connected assistants for real-time collaboration.
• Ensure onboarding and billing flows function smoothly.

Skills

React

TypeScript

TanStack

Zustand

TailwindCSS

IndexedDB

WebSockets

Cloudflare

Postgres

Tools

React

TypeScript

TanStack

Zustand

TailwindCSS

IndexedDB

WebSockets

Cloudflare

Postgres

Job description

We're building an AI assistant that does work for you, not just answers questions. You connect it to the tools you already use (Gmail, Calendar, Slack, Linear, Notion, and a few thousand others) and it runs in the cloud on its own computer, 24/7, whether or not you're online.

In practice that looks like: a briefing in your inbox every morning built from your calendar and email. Replies drafted in your voice, waiting for approval. A Linear ticket filed from a Slack thread. When it notices you asking for the same thing repeatedly, it turns that into a routine that runs on a schedule, or builds a small app with its own UI and database that you (and it) can use going forward.

You talk to it wherever you already are: our web and iOS app, iMessage, Slack, email, or inside Superhuman. Each person has their own assistant with its own memory of how they work, and assistants on the same team can talk to each other, so you can ask a colleague's assistant what they know about a project without pinging the colleague.

Our customers are founders and people who run a function at companies of 1 to 100, the ones who spend their day gluing four tools together by hand. We're a team of about 20, based in the Presidio in San Francisco.

About the role

You'll own the product surface of our assistant: the interfaces people use to hand work to it, watch it happen, trust the result, and come back tomorrow.

This is a product role with a frontend center of gravity. You'll follow problems wherever they go: interaction design, client architecture, the API route, the DB migration, the analytics event that tells us whether it worked. We work quickly. We need someone who can take an idea and turn it into reality: start from a rough direction or a Slack thread, sketch the interaction, ship a first version, and sharpen it from feedback. The PRD and the Figma file, when they exist, come after your first cut.

What you'll work on

Delegation primitives - handing work to the assistant, following progress, interrupting, approving, and undoing, for work that runs from seconds to overnight.

Memory made visible - the surfaces where users see what the assistant has learned about them, correct it, and watch it sharpen from every approval, edit, and rejection.

Routines and apps - the surfaces where users ask for real software (its own backend, data, and URL), install it, and monitor it and the routines that keep running after the conversation ends.

Connected assistants - shared sessions where several people and one assistant work in the same thread in real time, and the door between one person's assistant and a colleague's.

One assistant, every surface - the same assistant, memory, and identity whether someone opens the web app, texts it from their phone, or @-mentions it in Slack, Superhuman, or email. Includes onboarding that can start from a single text message.

Performance as a feature - a local-first client (TanStack Query + IndexedDB persistence) that opens instantly on a 10,000-message history.

First-time experience and billing - the funnel from landing page to trial to paying team, including the parts that make people trust a teammate with their credit card and their inbox.

What we're looking for

You've shipped complex web products and owned them after launch. We want to see the work: multiplayer experiences, large-data apps with real fetching and caching strategy, streaming or realtime UIs.

You have product taste and act on it. Give you a rough direction and you'll sketch the interaction, build it, and defend the tradeoffs.

You know React deeply: state management, when to lift into context or a store, what re-renders and why, how to write components other people want to reuse.

You're comfortable owning a full-stack feature end to end: migration, route, fetch, UI, analytics.

You care about performance and can explain your caching, invalidation, and optimistic-update choices. Duplicated network requests bother you.

You care about craft: loading states, empty states, error copy, keyboard shortcuts, the thousand small decisions that make software feel good.

You use AI heavily in your own workflow and have opinions about where it helps and where it doesn't. Professional AI-product experience is not required, but you're already probing current models and forming views.

You're autonomous and low ego. You don't wait for a spec, you ship, you get feedback, you iterate. You'd rather own five things fully than touch fifty.

We care about trajectory over tenure. Tell us what you built and how fast you got here.

Stack: TypeScript, React, TanStack (Query, Router, DB), Zustand, TailwindCSS, IndexedDB, WebSockets over Cloudflare Durable Objects, Cloudflare Workers, Postgres (Neon). Fundamentals matter more than a perfect match.

Bonus:

Built UIs for LLM-based products: streaming, uncertainty, progressive disclosure, non-determinism as a design constraint.

Real-time collaboration (CRDTs, Y.js, Automerge, or a homegrown equivalent).

Local-first architectures and sync strategies.

Design background or a strong visual sense; you can make it look good without a designer.

Founding or early frontend engineer who set standards from scratch.

Logistics

Location: San Francisco (Presidio) - in-person, 5 days a week in the office.

You must be based in the Bay Area or willing to relocate before starting.

US work authorisation required; open to O-1 or J-1 visa sponsorship for exceptional candidates.

Compensation: $180K-$240K base + meaningful early-stage equity, plus health, dental, 401(k), generous PTO, gym budget, and lunch.

Our process

Exceptional candidates go from first touch to offer within two weeks.

Application - resume, links, and one thing you've shipped that you're proud of. A short demo video or a link to the live thing beats a description.

System design interview - we walk through a real product problem together, typically a take-home slide with explicit constraints that you present back.

Final conversation - your questions, and we scope the work trial.

Paid work trial - in person, a few days to two weeks. You own a real feature end to end, present it at week review, and both sides decide.