- 🔭 open to Data Science / ML / AI roles
- 🌱 Currently sharpening classical ML fundamentals, SQL, and PyTorch
- 💬 Ask me about CNNs, RAG pipelines, or agentic AI systems
Languages
Machine Learning & Deep Learning
-FF6F61?style=for-the-badge){kind=icon}
LLMs & Agentic AI
Data Analysis & Visualization
Databases & Tools
A local RAG pipeline that extracts text, tables (pdfplumber), and images (PyMuPDF) from PDFs, captions diagrams with a local vision model (Llava via Ollama), and answers questions with page-level source citations. Uses two-stage retrieval — a bi-encoder for candidate generation and a cross-encoder for re-ranking — before passing context to a local LLM through Ollama.
Python FastAPI Streamlit LangChain ChromaDB Ollama
Identifies aquatic species from images by comparing them against 18,000+ pre-computed CLIP (ViT-B/32) embeddings across 20+ species. A multi-step agentic pipeline (plan → execute → reflect → adapt → self-validate) is run by a local LLM ReAct controller (Llama 3.2 via Ollama), with a rule-based fallback agent if Ollama is unavailable. FastAPI backend with a JS frontend.
Python FastAPI CLIP (ViT-B/32) Ollama (Llama 3) PyTorch
A hybrid deep learning pipeline combining a MobileNet CNN for feature extraction with a 25×25 Self-Organizing Feature Map for unsupervised clustering, processing 18,000+ images. Applies U-Matrix visualization to map cluster density and flag anomalous data points.
Python TensorFlow MobileNet CNN SOFM
Thanks for stopping by — feel free to explore my repos or reach out!