Data Scientist and Machine Learning Engineer

Designed and implemented a Python-based LLM evaluation framework with structured test datasets and A/B prompt experimentation, improving model reliability by 15% and reducing customer-facing errors. Developed reusable prompt templates, validation checks, and AWS monitoring dashboards, enabling systematic tracking of model experiments via GitHub and improving detection of hallucination and response inconsistencies.

Designed automated data pipelines using Python, AWS S3, Lambda, Athena, and Glue and built real-time dashboards in Power BI, achieving 100% data consistency and reducing debugging time by 30%. Designed and implemented a gamification algorithm (XP, streak tracking) using Python, improving user engagement by 20%. Conducted A/B testing using statistical analysis in Python (SciPy, NumPy) to evaluate feature performance and improve event tracking accuracy, reducing reporting inconsistencies by 10%.

Automated SEC EDGAR data extraction pipelines using Python (Requests, BeautifulSoup, APIs, Pandas), building structured datasets for financial analysis and reporting. Developed LLM-based financial summarization workflows using OpenAI API and Python, generating automated insight reports that reduced manual analysis workload by 35% and achieved 95% stakeholder approval. Performed exploratory data analysis using Python (Pandas, NumPy, Matplotlib) and developed time-series forecasting models (Scikit-learn), improving distribution planning accuracy by 20% and increasing inventory efficiency by 15%. Designed predictive demand forecasting models to optimize pharmaceutical supply chains, improving resource allocation across 20 regions.

Developed ML-based document classification models (scikit-learn) to automatically tag financial documents, reducing manual sorting by 40% and improving document processing throughput by 30%. Built data pipelines for document ingestion, feature extraction, and model training integrating MongoDB, MariaDB, and AWS storage, enabling processing of 100K+ documents per month. Designed ML inference APIs using FastAPI and AWS Lambda enabling real-time document classification in loan processing systems, reducing latency by 25%. Implemented CI/CD pipelines using Docker, Kubernetes, and Jenkins to automate ML deployment and monitoring, reducing release cycles by 50%.