What to Expect

Dawn Industries builds software that watches industrial robot cells (PLC, robot, gripper, sensors), diagnoses faults with evidence an engineer can verify, and applies a fix only when a human approves it. You will own the model of the cell: the compilers that turn vendor engineering artifacts (controller programs, project exports, robot backups) into one structured, queryable representation, and the representation itself. Your output is what the AI layer reasons over, so its correctness and its provenance are the product's credibility.

We develop against real hardware in our own lab.

APPLICATION NOTE: No PLC, robotics, or manufacturing experience is required. Learning to read industrial programs deeply is part of the job.

What You'll Do

• The Big Picture: Build the compiler, and the representation it targets, for the programs that run the world's factories, compiled into one typed, queryable, provenance-carrying model that an AI can reason over.
• Format Parsing: Build parser front-ends for vendor engineering artifacts, each against real test projects: Siemens TIA Portal sources and Openness/SimaticML XML exports, Rockwell L5X/ACD, Mitsubishi GX Works projects, and robot controller backups (FANUC TP/LS listings, I/O configuration, registers, alarm history), with more robot languages to follow (ABB RAPID, KUKA KRL, URScript).
• Program Analysis: Parse and analyze IEC 61131-3 logic (SCL/ST text, LAD/FBD networks) to recover behavior: state machines and their transitions, fault raise and clear conditions, timers, guard expressions, and a complete read/write cross-reference.
• Representation Design: Own the schema: a typed graph of tags, steps, faults, interlocks, and cross-device handshakes that stays general across vendors.
• Deterministic Compilation: Keep builds byte-stable and re-runnable: canonical serialization, content-addressed versions, and diffs that distinguish a changed fact.
• Corpus Validation: Prove generality against a corpus of hundreds of real-world projects across the major PLC and robot ecosystems, and build the query surfaces downstream consumers, human and AI, use to interrogate the model.
• Domain Fluency: Become a strong reader of PLC and robot programs, and encode cell semantics (scan cycles, latching, edge triggers, handshake protocols) into the model rather than into tribal knowledge.

What You'll Bring

• Degree in Computer Science, Computer Engineering, or equivalent practical experience.
• 3+ years of systems software engineering, or a portfolio that proves the same depth.
• Expert-level Python and real parser or compiler front-end experience: lexers and parsers, ASTs or IRs, graph data structures.
• Proven file-format reverse-engineering you can walk us through: a proprietary format you decoded and made legible.
• Data-modeling: schema design, closed vocabularies, versioning discipline, canonical serialization, content hashing.
• Not required: PLC, robotics, or manufacturing experience. Learning to read industrial programs deeply is part of the job.

Nice to Have

• You have worked on code intelligence at scale: cross-reference indexes, symbol graphs, code search, or the machinery behind an IDE's navigation.
• You have shipped a static analyzer, type checker, or compiler pass that found real bugs in other people's code.
• You have written a grammar, language server, or decompiler for a language.
• You have made a proprietary or abandoned format readable again: an office-document dialect, a game's asset files, a firmware image, a dead tool's project files.
• You have kept a large build hermetic: reproducible outputs and provenance you can audit.
• Familiarity with IEC 61131-3 environments (TIA Portal, Studio 5000, GX Works) or industrial robot controllers.

About the company

Dawn Industries

We're giving the world's factories one language for every robot they own.

Walk into any modern factory and you'll find something strange. The robots that weld your car, assemble your dishwasher, and lift things no human should lift are marvels of engineering, and they're still programmed like it's 1995. An engineer stands beside a two-ton arm holding a teach pendant, nudging it point by point, usually on a weekend, because that's the only time the line can afford to stop. Every robot brand speaks its own language. None of their tools talk to each other. The machines that build the physical world run on workflows nobody has seriously challenged in thirty years.

We think that's absurd. We're fixing it.

What we do

Dawn Industries builds VIM, the platform where robots from every vendor finally live in one workspace: programmed with the help of AI, tested in simulation before anything touches the floor, and kept in sync with reality for the entire life of the production line. VIM is the only software that can deploy and manage industrial robots 10x faster and cheaper.

The prize is enormous. Manufacturing is a multi-trillion-dollar slice of the global economy running on the least modern software stack in tech. Whoever rebuilds that layer changes how everything gets made. We intend to be that company.

Who we are

We're a small, founder-led team backed by Y Combinator. Our founders came to this problem obsessed with a simple observation: the hardest problems in robotics aren't in the lab anymore, they're on the factory floor, and almost nobody in software is paying attention. We spend as much time in plants talking to the engineers who keep production lines alive as we do writing code, and it shows in everything we build.

What it's like here

You'll work closely with the founders with real ownership over real problems. You'll talk to the people who actually run factories, and you'll watch your work land on production floors where the stakes are physical and the feedback is immediate.

If you are ready to push the frontier of AI in the physical world, come build with us.