My path in tech has been a fun and evolving adventure! For years, I built strong foundations in backend logic and problem-solving as a Java trainer and PHP web developer, followed by a deep dive into system troubleshooting as a senior technical support engineer.
Today, I am fully focused on Data Science, Machine Learning, and Generative AI. Because of my developer and support background, I don't just build models in isolated Jupyter notebooks—I care about clean architecture, handling real-world messy data, and deploying end-to-end applications that actually work in production.
(Click to expand the details of my recent builds!)
🕵️♂️ Fraud Detection Paysphere
An end-to-end fraud detection application designed to handle highly imbalanced datasets.
- The Challenge: Real-world fraud data is inherently skewed.
- The Solution: I engineered a robust pipeline using XGBoost and deployed it via Streamlit. The model was trained on 50,000 transaction records and successfully navigated a realistic ~10% fraud rate, maximizing precision without sacrificing recall.
⚖️ Legal Eagle (RAG Query Engine)
An intelligent document retrieval system for complex legal texts.
- The Tech: Built using LangChain and HuggingFace embeddings.
- The Solution: Implemented a Retrieval-Augmented Generation (RAG) architecture that allows users to seamlessly query, retrieve, and analyze dense legal documentation with high accuracy.
📈 FinSight Pro
An automated financial analysis platform and interactive dashboard.
- The Tech: Streamlit, NewsAPI, and Python data libraries.
- The Solution: Blends traditional technical stock analysis with live news sentiment analysis, giving a holistic view of market movements in one sleek web app.
⚾ "Moneyball" Scouting & Churn Analytics
Predictive pipelines solving specific business and sports analytics problems.
- The Solution: Utilized machine learning classification and regression techniques to predict customer churn, as well as applying advanced data visualizations (Seaborn, Matplotlib, PowerBI) for sports scouting analysis.
I love experimenting with new tech, but here is my daily driver stack:
| Category | Technologies |
|---|---|
| Languages |     |
| Data & ML |    |
| Deep Learning & Gen AI |    |
| Big Data & Web |    |
| Dev Tools |   |
💡 Current Obsession: I'm currently a massive fan of building and experimenting with cutting-edge open-source models, specifically utilizing
meta-llama/llama-4-scout-17b-16e-instructfor complex reasoning and development tasks!
Whether you want to discuss machine learning, collaborate on an open-source project, or just chat about cool data visualizations, my inbox is open!
- 📊 Kaggle: pankajkapri
- 💼 LinkedIn: Pankaj Kapri
⭐ Constantly building, optimizing, and learning.