A showcase of AI systems, production platforms, and open-source tools.
I'm a firm believer in learning by doing. This is a collection of my work, ranging from enterprise-grade SaaS platforms to experimental "from scratch" implementations.
AI-powered semantic document search platform using transformer embeddings and FAISS. Goes beyond keywords to understand meaning and context in your documents.
Enterprise multi-tenant SaaS platform for automated insurance loss run processing. Combines OCR and domain-specific LLMs to extract structured data from unstructured documents.
Production-ready recommendation engine using TF-IDF, SVD, and content-based filtering. Handles 930K+ movies with intelligent quality filtering, advanced features, and Django deployment.
Flask-based speech annotation platform for reviewing and correcting ASR transcripts with background jobs, row locking, and real-time progress tracking.
Educational repository implementing 18 essential ML algorithms from scratch using Python and NumPy. Features 2,500+ lines of documentation with mathematical foundations and production-quality code.
A simple, local-first habit tracker built with Flask and SQLite. Runs on your machine, keeps your data private. Features a monthly tracking grid, dashboard with charts, and Excel export.
A self-hosted audio search engine that identifies any audio clip in seconds. Uses spectral peak extraction and combinatorial hashing inspired by Shazam, running entirely locally.
An AI-powered web app that takes meeting recordings, transcribes them using multi-provider speech-to-text, and generates clean, structured Minutes of Meeting using LLMs.
A hand-crafted, dark-themed developer portfolio built from scratch with vanilla HTML, CSS, and JavaScript. Features cinematic landing, project showcase grid with pointer-tracked glow, and zero dependencies.
A web-based FC 25 player recommendation system using cosine similarity on 34 attributes. Features separate models for male and female players, multi-player radar chart comparisons, and advanced search filters.
A comprehensive, educational implementation of various autoencoder architectures for image processing - covering dimensionality reduction, reconstruction, denoising, similarity search, and image morphing.
A production-grade deep learning system for automated skin lesion classification using the HAM10000 dataset. Provides training, evaluation, and real-time inference for detecting seven types of skin lesions.
An open-source, interactive Python course that shows you what Python is actually doing under the hood - designed for developers who want to truly understand the language.
Interested in any of these projects or want to build something new together? I'm always excited to work on challenging problems.