Welcome to my GitHub profile! I genuinely enjoy coding and working on practical, end-to-end projects. I have experience working with a diverse tech stack and I like learning new tools and technologies by using them in real projects. I'm comfortable with different parts of building software and can pick up what I need to get the job done. I enjoy problem-solving through code and continuously learning by building and improving projects. Always open to collaborating on interesting technical work.
My portfolio: https://devadattap.github.io/
- 🎓 I'm currently pursuing an MTech in Data Science at IIT Palakkad. I completed my BE in CSE (AI & ML) with Honours in Data Science from FAMT under Mumbai University.
- 🤖 I'm especially interested in AI, machine learning, and data-driven systems.
- 🔌 I enjoy projects where hardware and software integrate.
- 🤝 I like working in teams and solving problems together.
- 🌱 I’m always trying to learn something new.
- 🐧 I shifted to Linux (Ubuntu) from Windows, and I am really enjoying it.
- 🛠️ Right now I am working on my MTech Project in collaboration with TCS research -- "Detecting citation manipulation and AI generated references in Computer Science literature."
While some of my repositories are private due to confidentiality, here's an overview of the types of projects I've worked on:
- Technologies: Python, Flask, C, libclang, PyYAML, isolate, Linux, CodeMirror
- Description: An automated grading system for C programming lab assignments — compiles and runs student submissions against instructor test cases in a sandboxed environment, with partial credit, static code checks, a live in-browser coding platform, and per-student and class-wide reports.
- Technologies: Next.js, TypeScript, PostgreSQL, Prisma, WebSockets, tRPC, Tailwind CSS, Auth.js
- Description: A real-time clipboard and notes sync app — content copied on one device appears instantly on others over WebSockets, with a markdown notes editor, room-code sharing with QR codes, guest access, and per-device management.
- Technologies: Python, Flask, PuLP, Linear Programming
- Description: An automated academic timetable generator for constrained scheduling scenarios like executive education programs, using integer linear programming to satisfy faculty and course constraints while optimizing session spacing. Supports Excel import/export and multiple schedule views.
- Technologies: Flutter, Dart, Supabase
- Description: A Flutter clinic management app built for my father's medical practice — stores patient records with full visit history, prescribed medicines, and billing, backed by Supabase for real-time, cloud-synced data.
- Technologies: Python, PyTorch, Vision Transformer, Graph Neural Networks, SymPy
- Description: A deep learning pipeline that converts handwritten mathematical expressions into LaTeX, combining a ViT/CNN image encoder with transformer and graph-neural-network decoders to handle nested symbols, 2D structure, and ambiguous handwriting, validated with SymPy.
- Technologies: Python, FastAPI, Sentence-BERT, Scikit-learn, NLP, Topic Modelling
- Description: An unsupervised NLP pipeline that classifies which ministry is being discussed in Indian parliamentary transcripts at the paragraph level, comparing seven methods — TF-IDF, semantic embeddings, topic modelling, and more — evaluated with LLM-generated weak-supervision references, served through a FastAPI web interface. Motivated from 'Sansad Bhashini' government initiative.
- Technologies: Python, Faster-Whisper, EasyOCR, Argos Translate, FFmpeg
- Description: A pair of real-time subtitle tools — one transcribes and translates system audio via Faster-Whisper, the other OCRs and translates on-screen video text via EasyOCR — both running multi-threaded pipelines with live overlay display and full offline capability.
- Technologies: Python, Playwright, Web Crawling, CLI
- Description: A CLI web crawler that generates sitemaps for sites missing one, rendering JavaScript-heavy pages and SPA routes with Playwright, normalizing URLs, scoping to a domain, and avoiding crawl loops. Outputs both sitemap.xml and plain-text formats.
- Technologies: Python, OpenCV, Computer Vision
- Description: A lightweight vehicle counting pipeline built entirely on classical computer vision — background-subtraction motion segmentation, blob extraction, centroid tracking, and route classification by exit zone — with no deep-learning object detector required.
- Technologies: Python, Selenium, Automation, Crypto Faucets, Testnet Wallets
- Description: A Selenium-based automation tool that claims daily cryptocurrency faucets across multiple wallets in parallel, reusing real browser profiles for saved logins and wallets, with retry/backoff logic and JSON-based state tracking to enforce per-wallet cooldowns.
- Technologies: Python, Reinforcement Learning, PPO, DQN, Simulation
- Description: A reinforcement learning framework that trains and benchmarks multiple RL algorithms (PPO, A2C, DQN, SAC, TD3, DDPG) against a realistic multi-floor elevator simulation, optimizing for waiting time, throughput, fairness, and energy efficiency against traditional rule-based control.
- Technologies: Python, Tkinter
- Description: A growing collection of desktop mini-games in Tkinter, each in its own module — currently a Wordle-style word-guessing game with color-coded feedback and a two-way number-guessing game, with hangman, sudoku, and more planned.
- Status: FAMT · TechVenture 2K24
- Technologies: Flutter, Dart, C
- Description: A campus treasure-hunt mobile app built for TechVenture 2K24 — teams solve C programming snippets to identify campus locations, scan QR codes to progress, with randomized cluster-based clue distribution and crash-safe time tracking.
- Technologies: Python, PyTorch, Transformers, DistilBERT, FLAN-T5, PyMuPDF
- Description: A document Q&A and summarization chatbot for PDFs, comparing an extractive DistilBERT QA pipeline against a FLAN-T5 + GloVe generative approach for more contextual, detailed answers.
- Technologies: Solidity, Ethereum, Truffle, React, Web3.js, Ganache
- Description: A decentralized insurance management app on Ethereum — Solidity smart contracts handle policyholders, insurers, terms, and claims, with a React + Web3.js frontend for filing claims and validating policies without intermediaries.
- Technologies: Python, Tkinter, NLTK
- Description: A GUI-based NLP toolkit for common text preprocessing tasks — tokenization, lemmatization, stemming, part-of-speech tagging, tense detection, clause segmentation, and voice transformation — built on NLTK for use without any command-line knowledge.
- Technologies: C#, .NET, WebAPI
- Description: A .NET Windows client-server platform for running live technical quiz contests — admins configure contests and participants, multiple quiz rounds including a buzzer-style round, and a real-time leaderboard. Not yet fully complete.
- Technologies: React, FastAPI, PostgreSQL, PostGIS, Leaflet.js, Python, REST API
- Description: A GIS-powered web and mobile platform for real-time monitoring of urban water supply networks — interactive network mapping, consumption and failure analytics, automated billing, and community grievance reporting so residents and authorities can track and respond to water issues together.
- Technologies: Python, Flask, OpenCV, YOLOv3, Docker
- Description: A real-time object detection and classification system using YOLOv3 on the COCO dataset's 80 classes, aimed at wildlife, agriculture, and environmental monitoring use cases, with a Flask web interface and Docker-based deployment.
- Technologies: Python, SMTP, Jenkins
- Description: A Python script that sends bulk emails to a recipient list read from file, wired into a Jenkins job for scheduled, hands-off distribution, using Gmail SMTP with app-password authentication.
- Technologies: Android, Java, SQLite
- Description: An Android app for boutique management — tracks clothing measurements, sizes, and order details per client, generates bills, and gives the shop owner a dashboard view across all clients, backed by a local SQLite database.
- Technologies: Python, OpenCV, Mediapipe, PyAutoGUI, NumPy, Computer Vision
- Description: A computer vision system that replaces physical mouse and keyboard input with hand gestures, tracked in real time via webcam. Uses Mediapipe's hand-landmark model to translate finger positions into system-level mouse and keyboard events. Patent filed.
- Technologies: Android, Java, Fuzzy Logic, SQLite
- Description: An Android application that diagnoses the likelihood of diabetes, anemia, and hypertension from user-reported symptoms using a fuzzy logic inference engine. Designed on the basis of https://doi.org/10.24203/ajfam.v8i2.6185
- Technologies: Python, OpenCV, Scikit-learn, Jupyter
- Description: A k-NN based character recognition system that segments characters from training images for interactive labelling, then recognizes characters in new images using the trained model, with visualizations of both training and recognition results.
- Technologies: Java, JavaFX
- Description: A JavaFX desktop app with five interactive mini-games — number guessing, birthday finding, mobile number finding, erased number finding, and the century game — built to make school-level math more engaging through gamification.
Feel free to explore my public repositories, connect on LinkedIn, or reach out via email. I'm always eager to learn, connect, and create innovative solutions. Let's build something amazing together!
Happy coding! 🚀