AI Engineer

What we're building

Frontier models now score above 170 on IQ tests. Reasoning isn't the bottleneck. Context is.

The context layer sits between an enterprise's siloed data and the agents that need to act on it. Stuff the context window and you trade quality for cost and latency. Use naive RAG and retrieval breaks the moment the question gets interesting. This gates most enterprise AI deployments we've seen, across private capital, professional services, edtech, and industrial data.

60x solves this. We built AI Brain, a knowledge graph platform engineered backwards from the agentic retrieval problem. Primary entity consolidation, chunk-level provenance, scheduled enrichment, Cypher over Apache AGE. Agents retrieve what they need and the surrounding context, no bloat, no hope-and-pray.

We run a Palantir model for workflows. The platform sits at the centre. Forward-deployed engineers wrap it around enterprise workflows we've templated. We retain each customisation as IP and feed it back into the platform, so each deployment gets faster, margins improve, and the moat widens. Same shape as Palantir's, different domain.

We're at the start. Clients include private capital firms, edtech, automotive data, professional services, and a growing list of global consultancies evaluating us against their internal GPT deployments. In the last two weeks, we shipped a redesigned ingestion pipeline, primary entity extraction with auto-enrichment, and an end-to-end SP500 demo across 500 companies. We move at this pace as a default.

The role

You'll be our second junior engineer, working directly with the CTO and a small senior engineering team on the parts of the platform that decide whether the context layer delivers.

• Ingestion and connectors. SharePoint, Google Drive, Gmail, DealCloud, and the next source on the list. Some clients hand us 400k+ files at 150+ GB and expect it to Just Work. You'll build the pipes and harden them.
• Knowledge graph internals. Primary entity consolidation, edge criteria, enrichment agents that decide when to call web search vs. internal tools, and the Cypher/Apache AGE query layer underneath.
• Agent infrastructure. LangGraph pipelines, Pydantic-typed state, prompt caching, the eval harness that keeps it honest.
• Product surface. The Next.js app where the graph, the reports, and the chat all meet the user.

You won't be boxed into a single layer. By month three we'd expect you to have shipped real work in both the Python backend and the TypeScript frontend, and to have opinions about both.

Our stack

• Frontend: Next.js (App Router), TypeScript, Tailwind, shadcn, deployed on Vercel
• Backend: FastAPI, Python 3.12, Pydantic everywhere
• Agents: LangGraph, Claude via Vertex AI, Gemini for cheap/fast tagging work
• Data: Postgres + Apache AGE (graph), moving toward AlloyDB Omni on GKE where it fits
• Infra: GCP - GKE, Cloud Run, Cloud SQL, Vertex AI, KMS
• Tooling: pnpm, Husky commit hooks (ruff, eslint, prettier, typecheck, test build, and an agentic check that fixes what it finds), Linear for issues, Claude Code as a daily driver
• We are opinionated about code quality and use AI coding agents hard. If pairing with Claude Code all day sounds uncomfortable, we're probably an odd fit.

What we're looking for

• Strong fundamentals in at least one of Python, JavaScript or TypeScript
• Something shipped, a project, a dissertation, an open-source contribution, a hackathon win, where you can walk us through architectural choices and what you'd do differently
• Comfort in an agent-native workflow. You write the spec, the agent writes the first draft, you review it. If you've never done this, prove you'll pick it up fast.
• Interest in knowledge graphs, retrieval systems, agent orchestration, or enterprise data engineering
• The taste and temperament to push back on a bad idea, including ours

You do not need:

• A CS degree
• Years of experience
• To already know LangGraph, Apache AGE, or any specific framework in our stack