Graduate Researcher — Diffusion Models at Queen Mary University of London

Developed and trained multimodal diffusion model on clean high-quality dataset using PyTorch on HPC clusters, achieving SSIM 0.90+ for high-fidelity speckle image generation. Conducted ablations on latent dims, noise schedules, and model design to balance technical depth and productivity.

AI/ML Computational Sr. Analyst at Accenture

Led research and production deployment of RAG systems, summarization, and QA on Azure with CI/CD, reducing processing time by 80% and operational costs by 40%. Fine-tuned GPT models for sentiment analysis, increasing production accuracy from 54% to 76%. Optimized code, prompts, and embeddings to improve output quality by ~75%. Collaborated with product teams to embed AI into applications.

AI/ML Engineer (Part-time) at Penta Group

Built end-to-end benchmarking workflows for large-scale LLMs (OpenAI, Gemini, Claude, Perplexity), automating semantic similarity and domain coverage evaluation for 4k+ outputs. Increased evaluation throughput by 80% and standardized internal research comparisons with reproducible metrics.

ML Engineer at Cognizant

Built and optimized PySpark large-scale ETL processes on Azure Databricks, improving data processing efficiency by 99% through schema enforcement and parallelization.