AI Engineer

LLMs suck at writing job posts.

If you understand why, we want to talk.

Our mission: build the best sales departments in the world and make elite coaching available to every team. It’s big. That’s where you come in.

We’re a fast-growing deep-tech startup backed by the investors behind Anthropic, TensorNet, and GROQ. In two weeks, we’re launching a self-serve product for mid-market—while continuing to scale wins with enterprise. Two motions. One team. Zero passengers.

• Own our insights engine: turn messy, human sales processes into measurable, coachable behaviors.
• Build the quantitative eval loop: design offline/online evals, define success metrics, wire them into CI to keep models honest.
• Instrument everything: trace, log, and attribute with Helicone; close the loop from prompt → outcome → revenue.
• Automate improvement: use LangGraph/LangChain to orchestrate tools, memory, and feedback for self-correcting flows.
• Ship weekly: your work will go live often and directly move client revenue.
• Partner with customers: jump into pilots with Sales/Success to see where the insights fail—and fix them fast.

• 3–7+ years in applied ML/ML-platforms, data-heavy backend, or agentic LLM systems (we care more about proof of shipping than years).
• You’ve built evaluation harnesses for LLMs or search/ranking systems and can explain trade-offs without hand-waving.
• Comfortable with Python/TypeScript, vector stores, embeddings, retrieval, and tool-calling.
• Strong grasp of observability & experimentation: tracing, A/Bs, guardrails, model/data versioning.
• Bias to simplify—you avoid Rube Goldberg pipelines and reach for the shortest path that survives production.

• Python, FastAPI, Postgres, Redis
• LangGraph/LangChain; OpenAI/Anthropic; RAG + function/tool calling
• Helicone for tracing; dbt/Airflow-style orchestration where it helps
• Infra: Docker, k8s, CI/CD on GitHub Actions

• Speed with safety: we ship small, instrumented changes; “fast” means measured.
• Coachability culture: we dogfood Yolk—every function runs on feedback and visible progress.
• High ownership: no ticket-takers. You’ll own problems end-to-end.

• Outsized surface area: a small team where your code changes how sales teams get coached—this quarter.
• Real momentum: enterprise logos in motion + imminent PLG launch.
• Career-defining work: you’ll do the best work of your career with people who expect that from themselves.

• Hybrid/Remote-friendly (US/EU time zones preferred)
• Competitive salary + meaningful equity + benefits

• Easy Apply here on LinkedIn or drop a brief note with links (GitHub, portfolio, most relevant project) and “why this problem” in 3–5 sentences.