Stephen A.

ML engineer with a master's in Machine Learning, currently doing applied research on post-training and evaluation for large language models. I work on making LLM systems reliable in production: structured/JSON output that actually parses, evaluation harnesses that catch regressions before users do, and fine-tuning pipelines for lower-resource languages where off-the-shelf approaches tend to fail. Also happy to take on data-quality work — I spent three years on OCR verification and document extraction, so I know what messy real-world text data looks like before it reaches a model. Open to short engagements and ongoing work.