Milind Shah
Head of Engineering at Brainbase Labs, leading development of autonomous browser agents and enterprise back-office automation systems. He has proposed a methodology achieving low cost and high accuracy through hardcoded coordinates and internal benchmarks.
Timeline (last 90 days)
Before adjusting prompts or model parameters, the Brianbase team first created an internal benchmark dataset containing questions and expected outputs together with Domino's non-technical business team.
Removing visual reasoning and hardcoding fixed X/Y screen coordinates for recurring UI elements was the optimization measure that produced the largest single accuracy improvement.
For teams with tight evaluation budgets, it is recommended not to make the initial evaluation too complex; deterministic evaluation criteria can check hallucinations, tool usage correctness, and adherence to the knowledge base at low inference cost.
Milind Shah's career experience includes software engineering roles at DoorDash and Adobe, growing a code data business from zero to eight-figure revenue at Merkle, and co-founding the evaluation platform Pipelines.
Milind Shah is generally optimistic about the AI revolution, but often thinks about the issue of skill degradation, and believes that a good AI engineer first needs a solid engineering foundation.
All Facts (5)
Before adjusting prompts or model parameters, the Brianbase team first created an internal benchmark dataset containing questions and expected outputs together with Domino's non-technical business team.
50%UnverifiedRemoving visual reasoning and hardcoding fixed X/Y screen coordinates for recurring UI elements was the optimization measure that produced the largest single accuracy improvement.
50%UnverifiedFor teams with tight evaluation budgets, it is recommended not to make the initial evaluation too complex; deterministic evaluation criteria can check hallucinations, tool usage correctness, and adherence to the knowledge base at low inference cost.
50%UnverifiedMilind Shah's career experience includes software engineering roles at DoorDash and Adobe, growing a code data business from zero to eight-figure revenue at Merkle, and co-founding the evaluation platform Pipelines.
50%UnverifiedMilind Shah is generally optimistic about the AI revolution, but often thinks about the issue of skill degradation, and believes that a good AI engineer first needs a solid engineering foundation.
50%