Pratik

Guha Roy

End-to-end invoice pipeline using OCR (Tesseract/pdfplumber) and Google Gemini for JSON extraction. Validation agent for GST compliance — GSTIN verification, CGST/SGST/IGST checks, duplicate detection. Integrated Supabase + Neo4j for storage and business relationship graphs. Auto-generates GSTR-1 and GSTR-3B returns.

CNN trained on the Lung and Colon cancer histopathological images dataset from Kaggle to classify CT scan images as cancerous or normal with high accuracy.

Dockerized FastAPI ML inference API classifying Iris flower species using a trained Decision Tree from scikit-learn. Features Pydantic validation, Swagger UI at /docs, and portable Docker deployment.

Robust HR Leave Management System and Client Meeting Scheduler using FastMCP — the Python framework for context-aware LLM-powered applications. Includes dynamic tools, prompt handling, and resources for a mock employee database.

Smart matching using semantic similarity (all-mpnet-base-v2 embeddings) and FAISS top-k search. LLM integration via Together.ai for response generation. FastAPI-first modular architecture.

Real-time facial landmark tracking using MediaPipe to monitor posture, attention, and fatigue. Audio feedback alerts for distraction/ergonomic issues. Interactive PyQt GUI with modular tracking, feedback, and UI components.

Smart assistant that reads WhatsApp messages, detects events using Gemini NLP, and syncs them with Google Calendar via OAuth. Eliminates manual deadline entry and boosts productivity.

Real-time computer vision tool using OpenCV and pre-trained deep learning models for age and gender classification via webcam feed.

Analysed job posting data to understand correlation between skills, location, company type, and salary. ML regression model with feature engineering and evaluation using scikit-learn, Seaborn, and Matplotlib.

Backend development for an AI-powered platform matching freelancers with suitable jobs. Designed scalable APIs and structured the codebase using clean architecture principles.

Won the hackathon organised by the Google Developer Student Club at Techno India University, West Bengal. Built and deployed a full-stack AI application from concept to production.

Secured 4th position out of 187 teams in this 24-hour nationwide hackathon at AJU Jharkhand. Built an intelligent automation solution under intense time pressure.

Completed and mastered the Postman API Fundamentals Student Expert certification, demonstrating proficiency in API design, testing, and documentation.

Completed the Microsoft Cloud Skill Challenge, acquiring hands-on knowledge in Azure cloud technologies, data fundamentals, and cloud infrastructure.