Omar Khattab

Assistant Professor at MIT EECS and a member of CSAIL (his own site, read 2026-08-20); the lead author behind dspy and the line of papers under it. His stated research question is the corner prompt-optimization occupies:

How do we program intelligent systems that are partly specified in natural language, that process natural language at scale, and whose quality and cost can be optimized using language models?

The work started at Stanford — Demonstrate-Search-Predict (arXiv 2212.14024, Dec 2022) with Santhanam, Li, Hall, Liang, Potts and Zaharia — and became DSPy a year later (arXiv 2310.03714, ICLR 2024). He also authored ColBERT and its late-interaction retrieval paradigm, which is where the @lateinteraction handle in DSPy's own release history comes from, and is credited on GEPA, the reflective prompt-evolution algorithm hermes-agent runs its self-improvement loop with.

His site calls DSPy "the first and most widely used declarative programming model for LLM systems." That is a first-party priority claim, and this corpus has not checked it against LMQL or the earlier templating libraries.

Related

dspy · prompt-optimization · stanford-university · hermes-agent · meaning-typed-programming