Teal started six years ago, when AI was barely usable. We bet on the future anyway. Built a company. Crossed five million users. Real revenue, real product, real trust with the people who use us.
We did all of that with the tools we had then. Now we get to do it with the tools we have now.
The original idea was always to give every ambitious person the kind of career infrastructure companies have always had. In 2020, that meant trackers, templates, and tools that helped people work the system harder. In 2026, it means agents that work the system for them.
The next Teal isn't a job-search tool. It's the AI that runs your career.
We're building agents that scout opportunities before they're posted. Agents that prep you for interviews. Agents that negotiate for you. Agents that remember every project you've ever shipped, so the next role finds you. Your career data, your career memory, your career vault. Owned by you. Traveling with you from job to job, for the rest of your career.
The closest analogy is what top athletes have always had. Agents, managers, coaches. A team that handles everything around the craft so the athlete can stay focused on the craft. Everyone else has been left to do all of that alone. We're closing that gap.
When ambitious people get this kind of leverage, the whole workforce levels up. More people in the right roles. More people growing into what they're good at. AI as the equalizer that makes raw talent the deciding factor again.
You'd be joining for what's next. The foundation is there. Five million users are there. The existing product continues to serve millions of people; the focus now is the new AI-native one.
You'd be one of three engineers on a small senior team, working directly with me (CEO), rebuilding the whole experience around agents that work for the user. You'll ship to millions of people who'll tell you, immediately, whether your thing worked.
These are real and unsolved here, and pretty much everywhere else. If you read these and want to argue with me about how you'd approach them, you're probably the person we want to talk to.
These are not toy problems. We don't have clean answers. We want someone who treats them as the actual job.
Edge-first AI infrastructure on Cloudflare. Workers, Durable Objects, AI Gateway, Vectorize, Queues, R2. Every user gets agents that run close to them, with state that survives across sessions, model routing we don't have to babysit, and embeddings that don't need a separate vector DB. This is one of the few production AI stacks built fully on the edge. It changes what's possible on latency and cost.
Multi-model agent orchestration. Anthropic, OpenAI, and whatever's next. Tool calling, structured outputs, streaming responses. Vercel AI SDK on the client. MCP servers for everything that talks to the world outside Teal. We treat the model layer as swappable on purpose. We want to be three weeks ahead of every model release, not three months behind.
The interface layer. React, TypeScript, shadcn/ui, Tailwind. Streaming everything. The agents have to feel alive. We obsess over time-to-first-token, retry behavior, graceful degradation when a model does something weird, and the moments of UX polish that turn a working product into one people actually love.
Career memory. The vault. Postgres, pgvector, structured and unstructured career data, semantic recall across years of someone's work. This is the layer that compounds. Every agent we ship gets smarter because the memory underneath it is richer than what any other career product has access to.
Evals. The gap we need you to close. We have agents in production making subjective calls. Is this resume bullet good? Is this interview answer strong? Is this the right next role for this person? We've been vibe-checking those answers and we know vibe-checking doesn't scale. Building the real system, harnesses, regression suites, latency and cost budgets, is one of the most important things the next senior engineer here will own. If you've felt the pain of shipping LLM features without evals, this is where you finally get to build the thing you wish you'd had.
Where the boring parts live. Rails, Postgres, the systems that have been earning their keep for six years and aren't going anywhere. The boring parts stay boring on purpose. The agent layer is where the new bets go.
This is what's in production today, not a commitment for what comes next. The new product isn't bound by these choices. A number of architectural decisions are still open, including agent state, retrieval, model orchestration, and evals. The next senior engineer will help shape them.
Bonus points: vector DBs in production · MCP servers · eval frameworks (Braintrust, Langfuse, homegrown) · experience in careers / HR / recruiting products
Apply with a GitHub repo, side project, or an AI-native thing you've shipped → founder chat with me → 1.5-hour max take-home coding exercise → review and discussion session → values chat with the team → paid project week where we ship something together.
Build what Teal was always going to become. Let's talk.
Commitment to Equal Employment Opportunity: (Come as you are. Feel welcome. Feel safe.) We are committed to safeguarding our workplace from all forms of discrimination and harassment on the basis of race, color, religion, sex, sexual orientation, gender identity or expression, age, national origin, disability, military status, or family status. This commitment extends to all aspects of the employment relationship, including recruiting, interviewing, selection, hiring, transfers, promotions, training, terminations, working conditions, compensation, and benefits.
