I'm a Computer Science (AI/ML) graduate focused on machine learning, deep learning, large language models, and applied NLP. My work spans experimentation, evaluation, embeddings and clustering, retrieval, model inference, APIs, and deployment.

I care about AI systems that are grounded, measurable, efficient, and useful beyond the demo.

• LLMs and NLP — few-shot learning, prompt design, structured extraction, taxonomy generation, RAG, grounding, and evaluation.
• Machine and deep learning — PyTorch, TensorFlow, transformers, embeddings, clustering, dimensionality reduction, and computer vision.
• Production AI — scalable inference, model optimization, asynchronous pipelines, vector search, Docker, and cloud deployment.

An AI-native production studio turning research papers and documentation into educational video, directed like a workspace rather than prompted like a chatbot. Eight departments coordinate through versioned artifacts, each running its own generate/evaluate/revise loop.

Document intelligence for classification, summarization, semantic search, and grounded Q&A with source citations.

FastAPI · React · Celery · PostgreSQL · pgvector · Redis · Docling

An agentic research assistant that checks every claim against the paragraph it cites, benchmarked for what that checking costs.

LangGraph · FastAPI · PostgreSQL · pgvector · Redis · LLM-as-judge · RAG evaluation

On-device waste classification for resource-constrained devices, built with React Native and TensorFlow Lite.

React Native · TensorFlow Lite · Computer Vision

• Developed self-service LLM workflows for knowledge extraction, hierarchical taxonomy generation, and data-grounded explanations over unstructured feedback.
• Combined few-shot prompting with embeddings, UMAP, HDBSCAN, BERTopic, and LLM-driven hierarchy induction.
• Optimized Amazon Bedrock inference and packaged the resulting pipeline for production deployment with Docker.

• Developed a knowledge-distilled deep-learning pipeline for real-time video sharpening on resource-constrained edge devices.

LLM evaluation · Taxonomy quality · Hierarchical clustering · Robust classification · Efficient inference

ML / Deep Learning — PyTorch · TensorFlow · Scikit-learn · Transformers · UMAP · HDBSCAN · BERTopic
LLM / NLP — Amazon Bedrock · OpenAI API · RAG · Embeddings · Few-shot learning · LLM evaluation
Engineering — Python · SQL · FastAPI · Celery · PostgreSQL · pgvector · Redis · Docker · AWS · GCP