Join Date
November 2025
Gender
Not Specified
Age
Not Specified
About Wyatt Marshall
2x early startup data/sw eng
Building a platform for environments, evals, and benchmarks to train and test computer agents
AMA about LLM/agent benchmarking and evals!
https://halluminate.ai/
Building a platform for environments, evals, and benchmarks to train and test computer agents
AMA about LLM/agent benchmarking and evals!
https://halluminate.ai/
Companies & Work
Halluminate
We help foundation model labs and enterprises train computer use AI with data and sandboxes.
A knowledge worker spends 90% of their day on a computer. A productive AI agent must similarly learn how to use our computers, browsers, and software to deliver real world value.
Today, browser and computer use AI is inaccurate, slow, and expensive. Model labs and enterprises who want to improve these agents are bottlenecked by two resources:
1) High quality datasets for benchmarking and evaluations
2) Realistic sandbox environments for safe and accurate testing/training (ex. a simulated version of Salesforce)
Halluminate offers a suite of products and services to address both these issues.
Our evaluation service combines proprietary datasets with high quality annotations to help our customers identify and prioritize the biggest failure modes of their computer- and browser- use AI.
Our platform provides a catalog of fully managed sandboxes, empowering our customers to safely and accurately test/train at scale, resulting in improved performance.
Our customers see
- Improved browser- and computer-use agent performance
- New and emergent frontier agent capabilities
- Data driven prioritization leading to exponentially faster development speed
- Increased revenue/sales via marketing from public benchmarks
Our paying customers already include leading computer use model labs and the two largest browser agent companies in the space.
Wyatt and Jerry are friends that met their first week of school studying CS at Cornell University. They’ve been working and studying together for 7+ years.
A knowledge worker spends 90% of their day on a computer. A productive AI agent must similarly learn how to use our computers, browsers, and software to deliver real world value.
Today, browser and computer use AI is inaccurate, slow, and expensive. Model labs and enterprises who want to improve these agents are bottlenecked by two resources:
1) High quality datasets for benchmarking and evaluations
2) Realistic sandbox environments for safe and accurate testing/training (ex. a simulated version of Salesforce)
Halluminate offers a suite of products and services to address both these issues.
Our evaluation service combines proprietary datasets with high quality annotations to help our customers identify and prioritize the biggest failure modes of their computer- and browser- use AI.
Our platform provides a catalog of fully managed sandboxes, empowering our customers to safely and accurately test/train at scale, resulting in improved performance.
Our customers see
- Improved browser- and computer-use agent performance
- New and emergent frontier agent capabilities
- Data driven prioritization leading to exponentially faster development speed
- Increased revenue/sales via marketing from public benchmarks
Our paying customers already include leading computer use model labs and the two largest browser agent companies in the space.
Wyatt and Jerry are friends that met their first week of school studying CS at Cornell University. They’ve been working and studying together for 7+ years.
