Harsh Kumar
Software Developer & AI/ML Engineer
Building production-grade applications and intelligent systems — from full-stack platforms to end-to-end ML pipelines.
About Me
I am a Full-Stack Developer and AI/ML Engineer currently pursuing my B.Tech in Electronics & Communication Engineering at IIIT Sri City (Nov 2022 – June 2026) with a CGPA of 7.5. I build production-grade applications and end-to-end machine learning pipelines that solve real-world problems.
Born in Aurangabad, Bihar, my journey has been shaped by a deep curiosity about how systems are built and how data drives intelligent decisions. As a Research Assistant at IIIT Sri City and a Data Science Intern at CodeClause, I have designed scalable backend services, deployed AI-based anomaly detection systems, and built collaborative filtering recommender systems. My focus spans full-stack development with Next.js, React, and Node.js to deep learning pipelines with PyTorch, LangChain, and MLflow.
Education
IIIT Sri City — B.Tech, ECE
CGPA: 7.5 • Nov 2022 – June 2026
4+
Years of Coding
10+
Projects Built
250+
Problems Solved
Harsh Kumar
@harshrajput4343
Technical Skills
Languages
Frontend
Backend & Databases
Data Science & ML
Deep Learning & LLMs
MLOps & DevOps
Data & Visualization
Core CS
Experience
Research Assistant — AI Backend Development
IIIT Sri City (June 2025 – Aug 2025)
- Designed and built production-grade Python services and RESTful APIs using FastAPI and Flask to deploy AI-based IoT anomaly detection systems.
- Implemented scalable data preprocessing and feature engineering pipelines using NumPy and Pandas for high-volume sensor and network data.
- Integrated CNN, attention-based deep learning models, and XGBoost into backend inference workflows for real-time prediction and monitoring.
- Developed MLflow-driven experiment tracking, model versioning, and validation pipelines to ensure reproducible and reliable model deployments.
Data Science Intern
CodeClause (Dec 2025 – Jan 2026)
- Configured AWS S3 buckets (raw/labeled/processed data), SageMaker Ground Truth labeling, 30+ annotated images.
- Deployed Lambda functions for preprocessing, configured IAM roles, orchestrated workflows via Step Functions.
- Built production-ready ML pipeline integrating S3, Lambda, Step Functions for end-to-end automation.
NumPy/Pandas
XGBoost
Detection
Tracking
raw / labeled / processed
Ground Truth
Orchestration
Featured Work
Production-grade applications and AI systems.
FlowLoG
Full-stack Trello-inspired Kanban board with drag-and-drop lists & cards, 37 Row-Level Security policies, checklists, labels, member management, dark/light themes, board templates, and real-time interactivity. 3-tier cloud architecture on Vercel + Render + Supabase.
MultiGPT
Multi-model AI chat app with smart auto-routing across 6 LLMs (Gemma, Nemotron, GLM). Real-time streaming, persistent chat history, tagging, full-text search, chat sharing via public links, glassmorphism UI, and Supabase Auth + RLS.
Quickart
Production-ready e-commerce platform with SSR, static generation, and incremental regeneration. Modular frontend with TypeScript, custom hooks, global state, and serverless backend with MongoDB, Clerk Auth, and Convex real-time cart sync.
ResoTrack
Serverless AI resume analyzer with React.js and TypeScript. Integrates Puter cloud for storage and authentication, Gemini API for ATS scoring, client-side PDF processing, and drag-and-drop file upload with real-time feedback.
AI Medical Chatbot
LLM-powered healthcare assistant using GPT-4 and Gemini with prompt engineering and RAG over Pinecone vector search. LangChain pipelines for document ingestion, semantic retrieval, and Flask-based inference APIs with caching and logging.
Book Recommender System
Collaborative filtering–based recommender using Scikit-learn and nearest-neighbor similarity on user–item rating matrices. Features threshold-based filtering (200 books / 50 ratings), Streamlit web app, and Docker + AWS deployment.
Waste Detection System
Production-grade waste detection using YOLOv5 with dataset annotation, preprocessing, augmentation, and real-time inference. Modular OpenCV pipeline with bounding-box filtering, confidence scoring, and Dockerized deployment on AWS and Azure.
Achievements & Leadership
Certifications
Oracle Cloud Infrastructure 2025 Certified Generative AI Professional
Complete Data Science, Machine Learning, Deep Learning, NLP Bootcamp
250+ Problems Solved
Leadership
Cricket Team Captain
Captained the college team in intercollegiate tournaments, leading match strategy and on-field performance.
Events Coordinator
Managed technical event logistics for 1,500+ participants across 15+ competitions, ensuring 100% on-time execution.
Let's Build the Future
Interested in collaboration or have a project in mind? Let's connect.