• 🎓 MSc Computer Science (Artificial Intelligence) — University of Kerala (2026 Passout) | CGPA: 8.75/10
• 🤖 Passionate about building end-to-end AI/ML pipelines, NLP applications, and production-ready data science solutions
• 💼 Hands-on experience through internships at Edunet Foundation and TechMindz
• 🧑‍🏫 Currently mentoring 15+ students in Python, OOP, and Data Structures at Transeduverse
• 📍 Based in Thiruvananthapuram, Kerala — available for immediate joining
• 🔭 Currently building production-grade AI systems with Docker, FastAPI, and LLMs

Python · Anthropic Claude API · Streamlit · Mermaid.js · GitHub REST API · 🌐 Live App

AI-powered tool that takes any public GitHub URL and instantly generates a plain-English explanation, structured architecture breakdown, data flow walkthrough, and a live Mermaid architecture diagram. Integrated Claude Sonnet API with real-time streaming output, analysing up to 20 source files per repository. Deployed on Streamlit Cloud.

Python · TensorFlow · Keras · Word2Vec · FastAPI · Docker · Docker Compose · 🐳 Docker Hub

Wide & Deep neural network for detecting vulnerabilities in Ethereum smart contracts, achieving 89% accuracy, 0.92 F1-score, and <4% false positive rate on the IJCAI 2020 dataset. Fully containerised with Docker Compose — FastAPI REST API on port 8001 and HTML web UI on port 8000. Published to Docker Hub; deployable with a single command.

docker pull deepuprakash/widennet
docker compose up

Python · FAISS · Sentence-Transformers · FLAN-T5 · Streamlit

Built a Retrieval-Augmented Generation (RAG) system for querying the Bharatiya Nyaya Sanhita. Implemented FAISS-based vector search with dense embeddings for semantic document retrieval and an interactive Streamlit UI.

Python · FLAN-T5 · Streamlit · NLP

End-to-end NLP application generating ATS-optimised resumes and cover letters using transformer-based semantic matching between job descriptions and candidate profiles.

Python · TF-IDF · Cosine Similarity

Content-based recommendation engine using TF-IDF vectorisation and cosine similarity to surface the top-N most relevant research documents.

• Built a production-ready AI-powered resume generator using FLAN-T5 with NLP-based semantic job description matching
• Designed structured prompt pipelines using Python and OOP best practices for automated document generation
• Deployed a Streamlit web app with automated PDF export; documented full pipeline for stakeholder review

• Performed end-to-end EDA and feature engineering on structured datasets with 10,000+ records using Python and Pandas
• Built and evaluated classification models using Scikit-learn; analysed precision, recall, and F1-score to select optimal models
• Communicated findings through clear stakeholder reports

• Delivering structured training in Python, OOP, and Data Structures to 15+ students
• Mentoring students on problem-solving and algorithmic thinking

• Designed a hybrid Wide & Deep neural network (TensorFlow/Keras) to classify Ethereum smart contract vulnerabilities
• Achieved 89% accuracy and 0.92 F1-score with false positive rate below 4% on the IJCAI 2020 dataset
• Built a full ETL pipeline: Solidity source → bytecode → opcode extraction → Word2Vec embeddings → model training
• Containerised and published as a multi-service Docker application to Docker Hub

• 🏅 Business Etiquette — TCS (Batch 04) · Score: 90% · 2026
• 🏅 Python for Data Science — NDZ Technologies (Dec 2023)
• 🏅 Personality Development Certificate · 2024

⭐ Results-driven. Always learning. Always building. ⭐