Sandipan Roy

Dynamic AI Engineer with a decade of extensive experience in the realms of Machine Learning, Natural Language Processing, and Generative AI. Aiming to leverage expertise in developing innovative AI solutions and leading technical teams within fast-paced environments, particularly in the technology and industrial sectors.

Senior Data Scientist | AI Architect at AlgoAnalytics (2020-01 – Present)

• Architected and deployed modular enterprise LLM pipelines with LangGraph, optimizing application tracking efficiency and throughput for IT enterprises.
• Collaborated with cross-functional teams to deploy quantized custom LLM weights on AWS Lambda and EC2, ensuring scalable, production-ready solutions.
• Led and mentored a team on rewriting, re-ranking, hybrid search, routing, and prompt optimization in advanced RAG systems.
• Directed development of multi-format retrieval pipelines (PDF, text, SQL, CSV) using LangChain, enhancing data accessibility and intelligence.
• Built a no-code ML/DL platform, enabling non-technical users to create AI solutions as a SaaS product.
• Implemented clinical entity extraction pipelines from EHR documents using Spacy, BERT, and Python, improving structured data availability.
• Boosted Predictive Model Accuracy by 30%: Leveraged advanced RAG techniques—including routing and re-ranking—to significantly enhance model performance.
• Cut ML Workflow Computation Time by 50–75%: Optimized data pipelines and processing strategies, dramatically accelerating end-to-end machine learning workflows.
• Reduced Manual Review Workload by 50%: Implemented generative AI models to automatically parse and screen resumes, streamlining candidate selection for ML/AI interviews.
• Enhanced Industrial Anomaly Detection with Fire & Smoke Alerts: Successfully integrated real-time fire and smoke detection into an existing computer vision anomaly detection system, improving safety and responsiveness.

Associate Data Scientist at AlgoAnalytics (2018-01 – 2019-12)

• Generated multilingual news headlines leveraging unsupervised NLP models in English and Japanese.
• Analyzed customer behavioral data to optimize recommendation engines using Spark and Hive.
• Managed the ML lifecycle with MLflow, accelerating model deployment by 40% through streamlined workflows.
• Conducted cloud-based model tuning and optimization, ensuring scalable, reliable analytics pipelines.

Application Developer at IBM (2014-01 – 2016-09)

• Managed and monitored real-time transactions within a Webmethods B2B environment, ensuring seamless operations and timely resolution of any issues that arose during transaction processing.
• Developed and implemented a new feature aimed at optimizing client communications, which resulted in a significant reduction in response times and improved overall client satisfaction.

MicroMasters® in Statistics & Data Science – Massachusetts Institute of Technology (2027)

B.Tech. in Electronics & Communication Engineering – West Bengal University of Technology (2013)