Summary
Behnam Asadi is an applied machine learning scientist with a decade of experience blending generative AI research and production validation, currently focused on RAG models and LLM agents at TD. He builds automated evaluation pipelines—LLM-as-a-Judge, embedding metrics, and synthetic prompt generators—and has shipped practical tooling like a PDF–Markdown round-trip evaluator to quantify document conversion fidelity. Concurrently a research associate at York University, he advanced image and text generation/compression (Deep Fourier Machines, transformer-based text compression) and achieved notable gains over JPEG and gzip baselines. Comfortable moving models from theory to deployable services (Dockerized crime-forecasting with a 7% test error), he pairs strong academic foundations from York and Sharif with hands-on engineering across computer vision, NLP, and time-series. Not obvious from the title: his work has produced sizable empirical improvements (e.g., 25.5 FID improvement and 2x text compression vs. gzip) that reflect a rare mix of signal-processing insight and pragmatic system-building.
11 years of coding experience
1 year of employment as a software developer
Abitur, ENGINEERING, Abitur, ENGINEERING at Leibniz Universität Hannover
Bachelor, Architecture, Bachelor, Architecture at Universität der Künste Berlin
German, English, Persian, French