Executed Project Canary, a large-scale study on reasoning failures in LLMs by designing 50+ multimodal experiments; improved evaluation depth and reduced hallucinations by 20% through engineering image-based reasoning tasks and refined system prompts. Collaborated on Project Orion, validating 200+ agent task-action pipelines, ensuring trace correctness and improved adherence to reasoning flows.

ML Research Assistant at University at Buffalo

December 1, 2024 - August 1, 2025

Architected an LLM-powered chatbot for AAC using Retrieval-Augmented Generation (RAG); fine-tuned a GPT-2 model on user-specific conversational history using LoRA and Hugging Face Transformers; improved personalized response accuracy by 30%. Developed a context-aware module integrating LangChain ChatBufferMemory and a predictive emoji recommender.

Associate Data Analyst at Oracle

June 1, 2023 - August 1, 2024

Engineered data pipelines to validate loan commitments and contracts across EMEA deployments; analyzed customer behavior patterns to identify anomalies and laid the foundation for an internal loan fraud detection algorithm; optimized contract lifecycle processes reducing manual verification overhead by 30%.

Data Scientist Intern at AB-InBev

June 1, 2022 - May 1, 2023

Led end-to-end development of the Empties Return Forecasting project for Africa Logistics Analytics; delivered a predictive model with 91% accuracy, improving forecast accuracy by 24%; enhanced on-time delivery by 20% and integrated forecasting insights into live logistics workflows across multiple markets.