Founding Engineer

Full-time. San Francisco, in person.

$120,000–$240,000 USD annual salary + approximately 0.5%–1.5% equity, depending on experience and scope.

Build the infrastructure for an autonomous company

We want to build a company where agents do substantial, useful work across engineering, support, research, sales, and operations. People set direction and make the decisions that need human judgment. Agents carry work forward, retain context, check results, and improve from feedback.

Screenpipe provides a foundation for that: open computer history. It captures screen and audio activity locally, makes it searchable, and gives agents context from what actually happened on a computer.

As a founding engineer, you will build both the product and the agent infrastructure we use to operate the company. You will work directly with the founder, use what you build in our own workflows, and turn the useful parts into capabilities other people and companies can rely on.

What you will own

• Agent infrastructure: persistent memory, context retrieval, tool integrations, MCP servers, orchestration, scheduled work, and handoffs between agents and people.
• Autonomous workflows that go from an event or goal to a verified result. Make them handle retries, interruptions, changing context, and partial failures without duplicating actions.
• Evaluation and observability: traces, reproducible failure cases, regression checks, and feedback loops that improve behavior while preserving permissions and human control.
• The core Screenpipe product: desktop capture, local storage, search, APIs, and interfaces that make computer history useful.
• Reliability on real machines: debugging crashes, memory and CPU usage, audio problems, and failures that are difficult to reproduce.
• Product ownership: talk to users, choose what matters, ship, and check whether the result helped. Keep the code understandable and maintainable as it grows.

Privacy, permissions, and data handling are part of the engineering work. You should know what an agent can read, what it can change, and when it needs a person to take over.

You already work heavily with AI

We want someone who pushes coding agents and models hard every day. You experiment with new models, run substantial workloads, build your own tools and automations, and keep refining your setup. You think about how much more useful work a small team could do with better agents.

Be ready to show your token usage, the systems behind it, and what that usage produces. We care about the relationship between tokens, cost, latency, and useful outcomes. A large token bill alone does not demonstrate engineering ability.

You should be able to:

• Show software or agent workflows you personally built and shipped, and explain the difficult decisions and results.
• Design systems that keep working beyond a successful demo, including memory, recovery, verification, and clear operating boundaries.
• Debug unfamiliar code and inspect AI-generated changes critically. You remain responsible for correctness.
• Move between product interfaces and backend or systems work. Our stack includes Rust, TypeScript, React, and Tauri. Depth in part of the stack and the ability to learn the rest matter.
• Take initiative, communicate directly, and change your mind when evidence warrants it.
• Work in person in San Francisco. Include your current location and when you could start here.

We assess demonstrated work and ownership. There is no fixed token-spend, degree, or years-of-experience threshold.

Show us how you work

Keep answers concise. Links, bullets, and public or anonymized examples are welcome.

1. Your AI usage: Roughly how many tokens do you use per day or week, across which models and tools? Separate personal usage from a team's total. If token counts are unavailable, share approximate spend and workload. What does that usage accomplish?
2. Your agent infrastructure: Walk us through your actual setup: coding agents, editors, memory, MCP servers, skills, orchestration, and recurring loops. What did you build or configure yourself? A sanitized AGENTS.md, CLAUDE.md, architecture diagram, or demo is welcome.
3. Work that runs without you: Show one workflow that continues beyond a single prompt. Explain its trigger, tools, state, output checks, failure recovery, and where a human takes over. What still breaks?
4. Things you shipped: Link two or three projects, with your exact contribution and an outcome you verified. Include one difficult bug or tradeoff and how you investigated it.
5. Learning and judgment: Roughly how many books did you read or listen to in the past year? Which two or three are your favorites, and what idea changed your work? Describe evidence or user feedback that changed a technical or product opinion.
6. Build with us: What would you automate first inside Screenpipe to move us toward an autonomous company, and how would you know it was working?

Do not include credentials, private prompts, or confidential customer or employer information.

Explore the project: https://github.com/screenpipe/screenpipe