Staff Parser/Compiler Research Engineer
dottxt (.txt)
Staff Parser/Compiler Research Engineer
dottxt (.txt)
Structured-generation platform for reliable LLM outputs
.txt builds structured generation: we make language models produce output that follows a given format, by deciding at every decoding step which tokens are allowed. We compile a JSON Schema, a regex, a grammar, a list of function definitions or a policy into that decision, fast enough to run inside the decoding loop. This is how .txt builds reliable and safe agents. We're looking for a developer to work on the parsers and compilers at the center of it.
Design and Develop: Craft high-performance parsers and data structures that integrate with LLMs.
Problem Solving: Research advanced parser and compiler technology and symbolic methods.
Writing Quality Code: Ensure the reliability and efficiency of our methods through rigorous testing and optimization.
Stay Ahead: Learn the latest, relevant LLM advancements.
You know Python at a professional level and have practical experience with "lower-level" languages (e.g. C/C++, Rust)
You have worked directly with parsers (e.g. Earley, LALR(1)), formal grammars, finite-state machines, and implemented efficient data structures. You are also capable of doing research and development in those areas.
You are familiar with our relevant technical report and open-source project.
You value simplicity, testing, and automation, and you can demonstrate this.
You communicate clearly, concisely, and consistently in written form.
🧭 The problem matters
Every agent that takes real actions needs a layer that guarantees it can only do what it was asked. The parser and compiler work in this role is what enables this guarantee.
🌐 Remote first
Work from anywhere in the world. We communicate mostly in writing and favor occasional organic discussions over recurring large team meetings.
💸 Competitive compensation and benefits
Market-rate salary plus stock options, health and dental insurance, and a 401(k) for US employees. We'll also get you a GPU if you don't have one.
Location
.txt is a fully remote company with the majority of the team in the US and EU.
For this role, you must be based in the US.
Please provide a 1-page resume in English.
Only apply to one position at a time.
Structured-generation platform for reliable LLM outputs
dottxt builds tools to make large language models act as reliable software components for enterprises by using structured generation that forces outputs into predefined formats like JSON schemas or regex. It works by applying constraints behind the scenes and validating results so outputs are predictable and easy to plug into existing systems, with an open-source Outlines library and enterprise-grade tooling. The company differentiates itself by combining statistical modeling and compiler technology to enforce strict output formats, plus early access from OpenAI and Cohere and a mix of community and paid offerings. Its goal is to make enterprise AI powered by LLMs predictable and scalable, turning generative outputs into deterministic data and services.
Explore more opportunities at dottxt (.txt)