AI/ML Engineer — MS in Artificial Intelligence @ WPI. I build production-scale Generative AI systems, multimodal ML pipelines, and LLM-powered applications. Open to new-grad AI/ML roles.
I work across computer vision, NLP, and end-to-end ML pipeline design.
WildSIFE — Species-invariant multimodal learning for captivity detection Frozen CLIP encoders + a shared feature extractor with adversarial Gradient Reversal training, so the model learns captivity cues (cages, enclosures) rather than species. 0.915 MCC in-distribution, 0.888 on cross-species OOD — beating Zero-shot CLIP, BioCLIP, and DINOv2 by up to 9.1% MCC. Under review at ECML PKDD 2026.
Chatbot with OpenAI Agents SDK — Multi-agent system 4 specialized agents (Document QA, Code Assistant, General Knowledge, Task Planner), 150ms avg response, 98% query-classification accuracy, 50+ concurrent users.
MVP Aptitude Automation — GPT-4o-mini question generator React + Supabase app for automated aptitude-question generation: 6 file formats, 98% parsing accuracy, 2–25 variants per question, CSV export with LaTeX support.
Languages: Python (advanced), JavaScript, SQL ML / AI: PyTorch, Hugging Face Transformers, CLIP, BLIP-2, scikit-learn, OpenAI Agents SDK, GPT-4 / Whisper Data & apps: pandas, NumPy, Streamlit, FastAPI, React, Supabase, n8n Infra: Git, Docker, SLURM (HPC)