About the role
HUD builds the infrastructure for training and evaluating AI agents. Our platform lets teams create reinforcement learning (RL) environments, benchmarks and evaluations, and our marketplace connects data vendors with frontier AI labs.
You will build and own our synthetic task pipeline, which generates the training tasks used to train AI agents.
Key responsibilities
Build systems that generate synthetic tasks and environments for agent training
Validate generated tasks for correctness, difficulty and solvability
Define and track metrics for diversity, realism and learnability of synthetic data
Run experiments to measure how synthetic data affects agent performance
Work with research engineers and data vendors to improve the pipeline
Requirements
Degree in Computer Science, Engineering or a related field, or equivalent practical experience
Solid software or machine learning engineering skills, shown through work, internships, research or projects
Strong Python, Docker and Linux skills
Experience with synthetic data, data generation or ML evaluation pipelines
Strong judgement on data quality and how to measure it
Good to have
Experience with large language models, RL or agent evaluations
Experience at an early-stage startup
Work location: On-site at our Singapore office, 9 Battery Road (Raffles Place). Full-time, permanent. 21 days of annual leave; CPF contributions for eligible employees.
We assess candidates on their skills and the work they can show us, such as projects, research, internships, open-source or previous jobs, rather than years of experience.
Requirements
- Degree in Computer Science, Engineering or a related field, or equivalent practical experience
- Solid software or machine learning engineering skills, shown through work, internships, research or projects
- Strong Python, Docker and Linux skills
- Experience with synthetic data, data generation or ML evaluation pipelines
- Strong judgement on data quality and how to measure it