Ship real AI products from week one, side by side with the founder.
Onsite · San Francisco Bay Area
About Solvea
We are a fully AI-native company. We use Claude Code, Codex, and agentic workflows to build useful, interesting products end to end: solvea.cx, voc.ai, smartcrawler.io, btcmind.ai, flatkey.ai, and more. More importantly, we are building a GTM Agent Swarm: specialized agents that research markets, find customers, write and review campaigns, run experiments, analyze voice-of-customer data, and turn product insight into distribution.
Why this is the moment
Join a fully AI-native company where Claude Code, Codex, and agentic workflows are how products get built every day.
Work in the room where decisions happen: this role is onsite in the Bay Area, side by side with the founder and core team, with no PM layer between the work and the call.
Build useful, interesting products end to end: solvea.cx, voc.ai, smartcrawler.io, btcmind.ai, flatkey.ai, and whatever we create next.
Help define the GTM Agent Swarm: agents for market research, customer discovery, campaign writing, campaign review, experimentation, VOC analysis, and distribution.
Set the taste bar for GTM agents: we want MKT builders and reviewers who can tell the difference between a working automation and an agent that is genuinely sharp, useful, and industry-leading.
Grow fast because the team is small and the cycle is early: we are hiring 10 people this cycle, and scope here is whatever you can hold, not whatever fits a training program.
The role
This is a builder role for new grads and early-career people who already make things and want to work on real AI from day one. You will help build AI-native products such as solvea.cx, voc.ai, smartcrawler.io, btcmind.ai, and flatkey.ai, while also building the GTM Agent Swarm we use ourselves. This role is onsite in the San Francisco Bay Area office because speed here comes from being in the same room as the founder and core team.
What you will do
Ship AI-native products end to end with Claude Code, Codex, and agentic workflows as part of the daily build loop.
Turn new product ideas into working systems fast enough that founder conversations become shipped product, not backlog.
Build and improve products across solvea.cx, voc.ai, smartcrawler.io, btcmind.ai, flatkey.ai, and new experiments we have not named yet.
Make the GTM Agent Swarm a real operating advantage, with agents that research markets, find customers, write and review campaigns, run experiments, and analyze VOC.
Bring exceptional taste to agent design: evaluate whether an agent feels useful, sharp, trustworthy, and good enough to become the industry bar.
Raise the pace and ambition of the team by proving that new grads here are operators from week one, not observers.
Create work that earns scope quickly because it works in production, not because it sounds good in review.
Who you are
You care about value creation and can point to things you built that people actually used.
You are deeply curious, self-directed, and you follow your own questions beyond the assignment.
You have imagination: you see non-obvious angles, reframe problems, and come back with better ideas than the obvious first pass.
You build end to end yourself and can walk through what you shipped, how it worked, and what broke.
You love GTM agents, MKT systems, or agentic workflows enough to have taste about them.
You can build, review, or sharply improve a marketing agent instead of merely describing one.
Judgment and shipping over slideware and process theater.
Credentials are not your story; evidence is.
AIQ bar
You use AI as a thinking and building partner every day, not as a wrapper around old habits.
You can turn Claude Code, Codex, prompts, tools, memory, and feedback loops into working systems.
You can judge AI output sharply: what is generic, what is useful, what is production-grade, and what should be rebuilt.
You compound fast because every project becomes a better prompt, skill, agent, rubric, or workflow for the next one.
Benefits
Unlimited AI tokens. Claude Code, Codex, and model usage are treated as core work infrastructure, not a rationed perk.
How we hire
You apply by doing a short, real build challenge; that is the application, and a resume is only a supplement. The process is our open-source recruiting agent, an async AI interview, then in-person conversations in our Bay Area office with the founder and core team. A human makes every call, and we move in days, not weeks.
What to send
A link to something real you built, ideally with a short note on how Claude Code, Codex, agents, memory, evals, or feedback loops changed the result.
A before/after artifact: what the first AI output or prototype looked like, what was weak, what you changed, and what improved.
A concise explanation of how you judged quality: what you rejected, what you kept, what shipped, and what you would rebuild next.
If you do not have a polished project, send a raw but original 24-hour build. We care more about AIQ evidence than presentation polish.
The challenge
Build something real. We use the challenge to see how you think, what you make, and whether you can turn an idea into a working system without waiting for permission. The strongest submissions are original, end-to-end, and clearly shaped around value creation — not polished decks, not copied boilerplate, not experiment theater.
Connect
Want to talk to the founder first? These work too.