Shakil Mosharrof

Shakil Mosharrof is a Machine Learning Engineer based in Dhaka, Bangladesh, with production experience across LLM systems, RAG pipelines, computer vision, and MLOps infrastructure.

He currently works at ADN Diginet within MetLife, where he built and deployed a recommendation system that increased lead collection and sales by 22%, and owns an end-to-end MLOps pipeline running across two geolocations on bare-metal servers. Prior to that he built production RAG pipelines for pharmaceutical companies at Delineate Pro (YC'25), designed computer vision APIs at Culture Hint, and engineered LLM-based recommender systems for a Japanese e-commerce platform at Gigalogy.

Earlier in his career he led the software and autonomy stack for MIST Mongol Barota, an autonomous rover that placed 1st at the University Rover Challenge competing against teams from 14 countries and secured over $50,000 in funding. That work shaped how he thinks about real-time inference, hardware constraints, and systems that have to work when failure is not an option.

He has published 7 research papers, including a best paper award at ICISET-2024 for a novel RAG architecture that combined vector retrieval, regex post-filtering, and T5 summarization to reduce hallucinations in LLM-generated outputs. He holds certifications from Microsoft, Databricks, NVIDIA, and Google across AI engineering, data science, MLOps, and cloud architecture.

His core stack is Python, PyTorch, LangChain, FastAPI, MLflow, Airflow, Docker, AWS, and Azure. He works across the full depth of ML systems, from model architecture through backend serving, infrastructure, and production debugging.