Hire Ruhin P. - Fine-Tuning Engineer

About ruhin

I ship ML systems end-to-end. Model training, evaluation harnesses, the C++ and Python plumbing underneath, the AWS services that serve it. The work I'm proudest of is usually the part nobody sees. At Yoloh, I built a production OCR + LLM pipeline (Llama-2 + GPT-4) that processes insurance documents at 95% F1 on a 10,000-document holdout. It automated 80% of field extraction and cut manual review from 250 to 100 hours per week. The pipeline that made that possible was the unglamorous part: rule-based validation, cross-field consistency checks, confidence-based human-in-the-loop QA. I also deployed the AWS Lambda microservice and DynamoDB + Neo4j backend that serve the resulting JSON, cutting p95 latency by 50%. At Stony Brook's Institute for Advanced Computational Science, I work on ML for scientific code generation. I fine-tuned a 7B CodeLlama model with QLoRA (rank 16, alpha 32) using a two-stage domain adaptation pipeline, improving generation accuracy on the MADNESS quantum chemistry codebase by 18%. To make that possible, I wrote the systems code around it: production C++ for HDF5 serialization with MPI-safe parallel I/O (merged to master), a Python tool server exposing 13 tools for HPC orchestration over SSH, a Tree-sitter ETL pipeline that cut dataset prep by 98%, and an evaluation harness scoring generated code on compilation, test pass rate, and API correctness. The pattern across both: the model is half the work. The other half is the data pipeline that feeds it, the evaluation that tells you it's actually working, and the infrastructure that lets it run reliably. I like all of it. B.Tech in ICT from PDEU (2022), M.S. Data Science from Stony Brook (2025), with internship stops at Blackcoffer and Yoloh in between. I came into ML through the engineering side — building things that work, then figuring out how to make them work better — and that's still how I operate. What I'm looking for: a role at a team shipping real LLM or applied ML systems, where the engineering is taken as seriously as the modeling. That could be ML Engineering, Software Engineering on an ML infra or platform team, or Data Engineering with ML depth. Open to early-career positions at companies where the work is technically serious. Core stack: Python · C++ · PyTorch · PEFT/QLoRA · AWS · Apache Spark · Docker · SQL Feel free to reach out or connect.

Common Questions

Other Fine-Tuning Engineers