The projects I build on my own time, and everything on GitHub.
An AI-powered SaaS that writes professional proposals for freelancers and agencies.
Full stack: a React front end, backend APIs and LLM integration with OpenAI and Claude, built for multi-user workflows, document generation and customization.
A multi-agent AI system that processes healthcare insurance claims end to end.
Orchestrator, extraction, validation, decision and explanation agents. Claim forms and medical reports are parsed with PyPDF2 and NLP, decisions carry confidence scores, and results come through a Streamlit front end with ReportLab PDF reports.
Agents built on LangChain's ReAct, Plan-and-Execute and Self-Ask patterns.
Works across OpenAI GPT, Claude and open-source LLMs, with FAISS and Chroma vector stores for retrieval, custom tools for API calls and memory for context.
Patient queues, doctor allocation and appointment flow for a healthcare practice.
Queue handling with real-time updates, and backend logic designed to stay low-latency under high demand.
PRIVATE REPOPredictive maintenance using ML, time-series analysis and optimization.
Scheduling with constraint programming and genetic algorithms, anomaly detection on IoT sensor data, and Streamlit dashboards for scheduling and analytics.
Scrapes product prices and exports them to Excel for monitoring.
Price monitoring with error handling; data exported with openpyxl and pandas.
Collaborative and content-based recommendations over scraped movie data.
Data scraped with BeautifulSoup and Selenium, models in scikit-learn, and trend analysis visualized with Matplotlib and Seaborn.
Sentiment classification for social media and product reviews, at 85–90% accuracy.
Naive Bayes, SVM, logistic regression and LSTM models, with text processing in NLTK, spaCy and TextBlob.