Founding Engineer

About the role

Founding Engineer

We’re looking for one engineer who can build. You’ll build the shared context layer for recursive companies: a data lake that ingests everything a team produces with AI, an ontology that turns it into trustworthy, typed knowledge, and the loop that lets people and agents improve from it. You’ll own real surface area from day one. We’re two people with years of runway, so you’ll have enormous autonomy and you’ll ship constantly. And because we’re a tiny team, the role is broad: alongside building, you’ll pick up internal ops, customer relationships, and whatever moves the company that week. The heart of it is building.

What "build" means here

- You have shipped something real: something with users, or something you use every day because you made it for yourself. You have gone from nothing to working software more than once.

- You’re comfortable across the stack. We ship native desktop apps on macOS, Windows, and Linux, a CLI and background daemon, agent integrations, and a cloud backend, across several languages. You do not need experience in all of these. You need to be the kind of person who picks up what they need.

- You reach for AI tools instinctively (Claude Code, Cursor, Codex) and you are faster because of it. If your first instinct under pressure is to stare at the code instead of handing it to an agent, this is not the right fit.

What you will work on

Core engineering

- Ingestion at scale: a data lake for heterogeneous sources, from ChatGPT and Claude Code transcripts to meeting recorders, notes apps, email, and full agent traces. Raw data first, no lossy summarization upfront. Cross-provider capture is still the hardest unsolved problem in the space.

- Ontology: the live state of a company, not a static schema. Live sources, derived fields, conflict resolution, and authority, so humans and agents can trust the answer and it stays current as new work flows in.

- Self-improving agents: skills as the unit of improvement. Sync a team’s skills across Codex and Claude Code, measure every invocation, mine traces for failures and repeated workflows, and propose reviewable edits back.

- Search and retrieval that makes thousands of sessions and documents queryable in milliseconds, across local and cloud.

- Human-AI coordination: team sharing, permissions, and policy. Hierarchical sharing models, separating what is shared from what is visible, and letting an organization share its thinking without oversharing. This is what turns Nessie from an individual tool into the context layer for a whole organization.

- Agent-facing surfaces: CLI, MCP, and plugins for every major harness, so any AI tool can reach accumulated human reasoning.

- Enterprise readiness: self-hosted and on-prem deployments, SOC 2, and the control plane behind them, as we move into larger and more regulated organizations.

A bit of everything

- We keep this part intentionally broad. On a two-person team, beyond shipping product you’ll help with internal ops, owning customer relationships, and the work that does not fit a job title. If you want to only write code and never touch anything else, this is not the role.