The Team Sergey Feldman

Data Cowboys operates as an R&D lab for AI research and evaluation, working with teams that need to figure out which models, methods, and technical bets are worth pursuing before committing resources to them. Founded in 2014 and based in Seattle with a presence in Chicago, the firm runs experiments, benchmarking, and failure analysis rather than shipping generic dashboards or one-off reports, and it delivers technical recommendations grounded in that testing.

The work spans several concrete tracks: data science and modeling that starts from messy, unstructured data and exploratory analysis; evaluation and benchmarking that digs into failure analysis for LLMs, search, ranking, and forecasting systems; and safety and reliability work including red-teaming, review, and reliability analysis. R&D sprints handle rapid exploration, baseline comparisons, and feasibility checks for teams deciding whether an approach is worth building out. LLM evaluation and guardrails engagements cover prompting, rubric design, reliability testing, and failure analysis, while search and ranking research looks at offline and online metrics, ranking experiments, and user-impact analysis. Forecasting and anomaly detection work involves comparing forecasting approaches and quantifying uncertainty rather than defaulting to a single model.

Sergey Feldman, a Principal Research Scientist at the Allen Institute for AI, and Ilya Barshai, a Principal Research Scientist at McGraw Hill, are among the people associated with the firm, giving the practice a research bench with direct experience inside applied AI labs. The agency describes itself as trusted by teams working on hard AI problems, which lines up with its focus on evaluation, benchmarking, and failure analysis rather than broader marketing or software delivery work.