Factory AI Interview Guide (2026): Droids, Agent Systems, and What to Expect

Factory was founded in 2023 by Matan Grinberg (CEO) and Eno Reyes (CTO), and it operates out of San Francisco. Grinberg came from a physics PhD track at UC Berkeley, and Reyes previously worked on deploying AI inside enterprises at Hugging Face. The product is a set of autonomous agents called Droids that work across the software lifecycle, from writing and reviewing code to running migrations and handling incident response, and they run from the terminal, an IDE, Slack, Linear, or a browser while staying model-agnostic. Factory raised a $15M Series A led by Sequoia in 2024, a $50M Series B in September 2025 (NEA, Sequoia, J.P. Morgan, Nvidia), and a $150M Series C led by Khosla Ventures in April 2026 that put the company at a $1.5B valuation, roughly $220M raised in total. Customers named publicly include Nvidia, Adobe, Morgan Stanley, EY, Adyen, and Palo Alto Networks, and the company runs a five-day in-office setup in San Francisco.

Process

Factory does not publish its interview loop, so treat this as a read based on the company’s stage and product rather than a fixed script. Public signal is thin: candidate reports point to a recruiter or hiring-manager screen followed by a technical phone round, and job posts describe coding, system design, and behavioral components. Expect something close to a recruiter screen, a practical coding round, one or two deeper sessions that mix system design with agent and LLM systems questions specific to the product, and a behavioral conversation that at this size often includes a founder. For a startup around 125 people that is hiring fast, plan for a loop that runs a few weeks rather than months, though a small team can move quickly when they see a strong match or stall when calendars are full. Roles vary a lot in emphasis: an AI engineer loop will weight LLM and agent design, while a product or platform role leans on full-stack or backend systems, so ask the recruiter what each stage covers and match your prep to it.

What they actually ask

• Design the architecture of a coding agent: how it plans a task, breaks it into steps, calls tools, and recovers when a step fails or a tool returns garbage.
• Set up retrieval over a large codebase (RAG) so an agent pulls the right files and symbols, and explain how you would chunk, index, and rank context.
• Evaluate an LLM that writes or reviews code: what metrics you would track, how you would build a test harness, and how you catch regressions when you swap models.
• Manage a limited context window under real latency budgets, deciding what to keep, summarize, or drop as an agent works through a long task.
• Build a backend service that orchestrates agent runs, including queueing, streaming output, timeouts, retries, and safe handling of tool side effects.
• A hands-on coding task judged on clean, correct structure, closer to real engineering than contest puzzles, so expect data structures and clear reasoning over trick questions.

Levels and comp (2026)

Factory does not post salary bands, and there is little public levels.fyi data on the company yet, so treat any number as directional. Packages at a startup like this pair a competitive San Francisco base with meaningful equity, and cash bonuses are usually small or absent. As a rough frame for a well-funded SF startup in 2026, senior engineer total comp often lands somewhere in the mid-$200Ks to mid-$300Ks between base and annualized equity, with entry and mid levels below that and staff or specialized AI roles higher, but you should confirm against levels.fyi and recent Blind threads rather than trusting a range here. On equity: the $1.5B valuation makes the paper value real, but it is private stock, so weigh strike price, vesting schedule, dilution across future rounds, and an unknown liquidity timeline before you compare offers. Ask directly about the share count, the current preferred price, and the option expiration terms, since those matter more than a headline percentage.

Prep priorities

Spend most of your time on agent and LLM systems, because that is the product. Be ready to whiteboard how a coding agent plans, uses tools, and recovers from failure, and to talk about retrieval over large codebases, evaluation harnesses for model output, and context management under latency limits. Have one backend or systems project you can defend end to end, including how you handled failures, retries, and observability. Keep your coding sharp on core data structures and clean implementation, since the hands-on round rewards correct, readable code over clever tricks. Because Factory ships an AI coding tool, using agentic coding tools yourself and having opinions about where they help or break will land well. For the behavioral and founder conversation, come with a specific answer for why autonomous software engineering and why Factory, plus questions that show you understand what it takes to run agents safely inside a real enterprise codebase.

Frequently Asked Questions

Do I need LLM research experience to get hired? Not for most roles. Factory hires AI engineers who apply LLMs in production, plus backend, platform, full-stack, and data engineers. Practical systems skill and a track record of shipping matter more than publishing papers, though the AI-focused roles do expect hands-on LLM work.

Is the coding round a LeetCode grind? Public signal is limited, but the framing points toward practical engineering over hard algorithm puzzles. Stay fluent with core data structures and write clean code, and expect problems that resemble real work more than contest questions.

How long does the process take? Expect a few weeks for a startup this size, though small teams can move fast for a strong match or slow down when schedules are tight. Ask the recruiter for the stage list and timeline up front so you are not guessing.

Is the job remote? No. Factory runs a five-day in-office setup in San Francisco, near Caltrain, so plan for on-site work if you apply.