Member of Technical Staff

Codos, Inc.; San Francisco (CA); On-site; USD 180,000 - 260,000; Full time

Benefits offered by this job

Competitive compensation and equity

No limit on AI tools: models, tokens

Offsites and business trips

Job summary

Codos, Inc. in San Francisco is building a company brain and the agents that run on it. You will ingest Slack, email, calls and documents into a permissioned knowledge graph and improve enterprise data retrieval.

You will deploy into client environments, including on‑prem, fix data issues, and turn forward-deployed problems into product with the CTO and a small, high-impact team.

Qualifications

• Experience building memory systems for organizations.
• Experience at Speedrun, YC or similar early-stage startup.
• Ability to allocate and manage token budgets.
• Shipped production LLM systems, such as knowledge graphs and retrieval pipelines.
• Own your work end-to-end: design, build, deploy, and fix when it breaks.

Responsibilities

• Ingest data from Slack, email, calls, and documents into a permissioned knowledge graph.
• Improve retrieval on enterprise data and advance our enterprise benchmark.
• Build the agent layer where agents run business functions on the brain.
• Deploy into client environments, including on-prem, and fix data-related issues.
• Turn field problems from forward-deployed engineers into product within days.

Skills

Memory systems

Startup experience

LLM systems

Multi-agent setups

End-to-end ownership

10x engineer

Job description

Build the company brain and the agents that run on it.

What you'll work on here

• Building the company brain: ingestion from Slack, email, calls and documents into a permissioned knowledge graph
• Improving retrieval on messy enterprise data and moving the number on our own enterprise retrieval benchmark
• Building the agent layer, where agents run real business functions on top of the brain
• Deploying into client environments, including on-prem, and fixing what a client's data breaks
• Turning problems our forward deployed engineers bring from the field into product within days

Ideal profile

• Built personal or corporate memory systems successfully.
• Worked at a Speedrun, YC or other early-stage startup.
• Can productively spend $30k in tokens per day.

What you bring to the table

• Pre-AI, you were a 10-100x engineer.
• You have shipped production LLM systems, such as knowledge graphs, retrieval pipelines, agent runtimes and evals
• You run your own multi-agent coding setup and ship a lot every day
• You own your work end to end: design, build, deployment, and the fix when it breaks

Why you should join our engineering team

You will work directly with Gleb, our CTO and a former machine learning engineer at Meta, as one of the first engineers on the core product.

Enterprise clients use what we ship every day. Our forward deployed engineers sit inside those clients and bring back real problems every week, so you always know whether what you built worked.

The product has two halves, the brain and the agents that run on it. Both are early enough that one engineer can own a whole piece.

Company Benefits

• Competitive compensation and equity
• No limit on AI tools: models, tokens and whatever your setup needs
• Offsites and business trips

We help large companies become self-evolving organizations.

• Real impact: We generate millions of dollars of impact for NASDAQ-listed and PE-backed clients, and we are going after the $500B services industry with software.
• Small team, huge impact: We are fewer than ten people, backed by a16z speedrun. The team is ex-McKinsey, ex-Meta ML and ex-Tesla, and the founders have run companies before: Dima scaled Everclear to $1B in monthly volume as CEO. Each new hire changes the company's trajectory.
• Client impact first, AI-native: We measure ourselves by the money a client saved or earned. Everyone here builds their own agents, and the company runs on its own company brain, the same way we teach clients to work.

Demand is ahead of our headcount, and we are hiring to close that gap.