I am an AI/ML systems developer and full-stack software engineer who bridges the gap between machine learning research and scalable production deployments.
Over the past year, I have completed 6 intensive engineering internships across AI research, full-stack architecture, and generative AI pipelines. I've engineered everything from AST-driven autonomous code analyzers (Top 10% National Hackathon) to cryptographic socket protocols with AES-256-GCM encryption and enterprise demand forecasting models.
My foundation combines deep mathematical rigor in statistical learning with high-throughput backend infrastructure in Python, Node.js, React, Docker, and Cloud architectures (Google Cloud, AWS).
Awarded Star Performer recognition for exceptional delivery. Architected and deployed 3 core projects: Nova Voice Assistant with real-time speech synthesis, a multithreaded TCP socket chat engine, and an interactive BMI telemetry application.
Built end-to-end machine learning and analytics engines: Real-Time Credit Card Fraud Detection (0.8886 ROC-AUC), Multi-Symptom Healthcare Disease Recommender, and IPL Cricket Match Telemetry Dashboard with interactive Plotly visualizations.
Developed production NLP and Vision microservices: Streamlit-based Groq LPU Llama 3.3 70B Assistant, Google Translate API Neural Multi-Language Translator (50+ languages), and In-Browser Computer Vision Gesture Recognition with TensorFlow.js.
Completed all 8 full-stack engineering milestones: Node.js + Express backend infrastructure, client/server validation pipelines, REST API architecture with MongoDB JWT authentication, live weather integration, and production caching middleware.
Shipped 5 AI modules in 30 days: Rule-Based RegEx NLP Chatbot, Unbeatable Minimax Tic-Tac-Toe AI, OpenCV Haar Cascade Face & Eye Detection, BLIP Transformer Image Captioning with PyTorch, and TF-IDF Movie Recommendation Engine with Cosine Similarity.
Engineered production Generative AI applications leveraging LLM orchestration, structured prompt engineering, vector database embeddings, and RAG document intelligence workflows for multi-agent systems.
Whether you have a technical opportunity at your company, an open-source project, or want to discuss machine learning architectures, I'm always open to connecting.