Build
Projects, essays, and notes on AI systems, evaluation, and the industry behind them.
Build
Systems thinking, shipped products, and the work of turning ambiguity into something real.
Evaluate
Model evaluation, quality, human judgment, and what useful measurement actually looks like.
Industry
Data labeling, incentives, AI operations, and where the role of data science is going.
Build
Systems, product thinking, and the work of shipping useful things.
What a Data Scientist Actually Does at an AI Company
The role is becoming less about isolated analysis and more about building systems that connect data, product, operations, and engineering.
From Fuzzy Problem to Shipped System
On turning ambiguous business problems into thoughtful, useful, shipped systems.
Why the Best AI Data Scientists Think Like Builders
The job is changing from analysis alone to building things that work in the wild.
Evaluate
On quality, judgment, and what useful evaluation really requires.
Model Evaluation Is More Than Benchmarks
Why good evaluation depends as much on design and judgment as on metrics.
What “Quality” Really Means in AI
A closer look at model quality, human judgment, and operational reality.
Human Judgment Is Part of the Model
The role of raters, disagreement, and evaluation design in real AI systems.
Industry
The human and operational systems behind modern AI.
Thoughts on the Data Labeling Industry
The hidden labor, incentives, and systems behind modern AI.
The Hidden Economics of Human-in-the-Loop AI
Why labor design, incentives, and operations shape model outcomes.
The Future Shape of Data Science in the AI Era
Why the role is shifting from analyst to builder.