Come build the thing that doesn't exist yet.

The vibe

You love 0→1. You crave the blank database, the first agent that reliably extracts truth from a stubborn government PDF, and the MVP that ships in three days. You build natively with AI and stay close to the customer.

• Some days you're deep in messy data.
• Some days you're standing up a new vertical.
• Some days you're rebuilding the website because it needs doing.

Whatever it takes to win.

What you'll own

• The Reasoning Layer (core): orchestrate the agentic system that turns raw, unstructured sources into structured, trustworthy, source-traceable output.
• The Knowledge Base: own chunking, embeddings, retrieval, and evaluation.
• The Engine: scrapers and pipelines against hostile sources — locked PDFs, government portals, legal databases.
• The Architecture & Infra: a global, multi-country Postgres schema from Day 1, and the deployment (AWS/GCP). If it breaks, you fix it.

Who you are

• Agentic builder (core): hands-on with an orchestration framework — LangChain, LangGraph, or LlamaIndex — on a major model (Anthropic, OpenAI, or Google). You've shipped agents that retrieve, call tools, and reason.
• Python-first: fluent and hands-on.
• Database management: Postgres (+ pgvector), a vector DB (Qdrant / Pinecone / Weaviate), and Redis.
• A Builder (4+ years): built APIs and systems from scratch (plus points for founder / employee 1–5). Pragmatic full-stack, backend-heavy, ships React when the customer needs it.
• Data & quality obsessed: deep scraping experience — wrong data in regulated products has real consequences.
• Founding DNA: priorities shift fast; you adapt.

How to apply

The founders read every application. In exchange, skip the résumé theater and show us how you actually build. Send an email to karina@kensai.ai with the exact subject line:

Kensai | Founding engineer

Applications without this subject line won't be read. Include these three things:

1. Proof you build. A repo, a shipped product, a scraper, or a schema you're proud of. Add one sentence on why it was hard, plus your LinkedIn profile.
2. A 20-minute take. Take Product A or Product B and sketch how you'd build a SKU-level intelligence dashboard for it (tracking claims, ingredients, sourcing, attributes). We don't want polished code — send a schema architecture, or a Loom walking through your approach. We want to see your product instincts.
3. One sharp question. The question you need answered before betting the next few years of your life on us. Skip Claude — we want to know what you are interested in.

The idea

We don't have this role written into a job spec yet. If our mission genuinely excites you, and you can see exactly where you'd create value here, let's meet.

Maybe you're a GTM operator, a regulatory wizard, or something we haven't thought of. Show us how you'd move the mission forward.

How to apply

Don't send a résumé. Share with us what you'd actually do. Email karina@kensai.ai with the subject line:

Kensai | The mission

Why this mission, what you'd own, and one concrete thing you'd build or fix in your first month. Add a link to something you've made and your LinkedIn.