Harsh Nerkar
Aspiring AI Engineer
An aspiring Artificial Intelligence engineer passionate about building intelligent, human-centered systems — from machine learning models to full-stack AI products. Driven to turn data into impact, and ideas into things that ship. Let's build something remarkable together.
Harsh Ramesh Nerkar
AI / ML Engineer • Nagpur, India
I'm Harsh — a final-year B.Tech Computer Science student with hands-on expertise in Machine Learning, Artificial Intelligence, and Computer Vision. Backed by three professional engineering internships, I build end-to-end, production-grade systems — from multi-agent AI architectures and RAG pipelines to real-time computer vision applications.
I work fluently with Python, Large Language Models, and modern data tooling, and I care deeply about engineering intelligent, scalable systems that drive measurable real-world impact — without losing sight of the user experience around the model.
Skills & Stack
Languages & Databases
Python
C
SQL
PostgreSQL
MySQL
MongoDBAI / ML Libraries
TensorFlow
Keras
PyTorch
OpenCV
MediaPipe
scikit-learn
Pandas
NumPy
MatplotlibTools & Platforms
Docker
Git
GitHub
Power BI
Tableau
N8N
Excel
Jupyter
CanvaWhere I've Worked
Agentic AI Engineer Intern
Architecting enterprise-grade Autonomous Multi-Agent Systems with multi-phase lifecycles (dynamic planning, self-reflection, and quality gating). Engineering resilient RAG pipelines and structured reasoning systems using modern LLMs (Gemini, Llama, Groq) with circuit breakers and HITL safety.
Machine Learning Engineer Intern
Architecting an end-to-end AI-powered Mental Wellbeing application: a multi-agent system on Llama 3 (Orchestrator, Coach, Safety, Memory) with a RAG pipeline over Neon PostgreSQL for context-aware, hallucination-resistant responses grounded in validated mental-health literature.
AI / ML Intern
Built an Indian Sign Language translator (CNN/RNN/LSTM + OpenCV) for real-time gesture-to-text, a Safety Helmet Detection system fine-tuning YOLOv8 with MediaPipe & OpenCV, a gesture-driven virtual keyboard, and a WhatsApp reminder bot.
Featured Projects
Python AI Engineer – Autonomous Agents
Enterprise autonomous document generation agent that transforms natural language business briefs into publication-ready Microsoft Word deliverables via a 4-phase reasoning lifecycle (Plan -> Execute -> Reflect -> Compile) with interactive FastAPI Web Studio.
- •Dynamic 6-stage TODO execution planner for ambiguous business briefs and conflicting departmental policies.
- •Self-check reflection audit engine detecting missing SLAs, milestones, and calculating objective quality scores.
CaptionAI Multimodal Studio
Bidirectional AI vision platform featuring Text-to-Image diffusion across 8 art styles, Vision-to-Text captioning with TTS audio narration, multimodal concept remixing, and an interactive Model Training Hub.
- •Bidirectional synthesis: Natural language prompt-to-image diffusion & image-to-multistyle captioning.
- •Integrated Text-to-Speech (TTS) audio narration for generated captions and alt-text tags.
Agentic AI Systems (Production SRE)
Production 20-Day Autonomous SRE & Multi-Agent Architecture featuring autonomous agent swarms, hybrid RAG, circuit breakers, and human-in-the-loop (HITL) safety boundaries.
- •Autonomous multi-agent swarms with distributed asynchronous task orchestration.
- •Hybrid dense-sparse vector RAG retrieval with verified hallucination resistance.
Indian Sign Language Translator
Real-time ISL gesture-recognition system combining MediaPipe hand tracking with a deep-learning classifier to detect Indian alphabets and numbers and render full words and sentences on-screen.
- •Real-time gesture-to-text using MediaPipe 21-point hand landmarker tracking.
- •Deep CNN classifier detecting 26 ISL alphabets and numbers with >95% accuracy.
Universal Text-to-SQL AI Generator
Multi-table relational schema-aware SQL generator powered by Gemini & Groq LLMs with automated execution testing, self-correction, and multi-dialect SQL support.
- •Dynamic schema introspection across PostgreSQL, MySQL, and SQLite databases.
- •Automated query syntax validation and iterative error self-correction.
Mental Health Wellbeing Companion
Research-backed AI platform for overthinking, stress, and anxiety — a multi-agent conversational system on Llama 3 with a RAG pipeline over peer-reviewed literature, stored in Neon PostgreSQL.
- •Multi-agent system on Llama 3 — Orchestrator, Coach, Safety & Memory agents.
- •RAG pipeline over Neon PostgreSQL grounded in validated mental-health literature.
Safety Helmet Detection
Construction-site safety-compliance system fine-tuning YOLOv8 with CNN/LSTM layers alongside MediaPipe and OpenCV to automatically detect missing safety helmets in real time.
- •Fine-tuned YOLOv8 for real-time PPE-compliance monitoring.
- •MediaPipe + OpenCV pipeline for construction-site footage.
Gesture Virtual Keyboard
Touchless text-input system enabling typing through real-time hand tracking — a gesture-driven virtual keyboard built with computer-vision hand-pose detection.
Education & Foundation
B.Tech Computer Science & Engineering
HSC (12th) General Science
SSC (10th) General
Verified Certificates
AI / ML Internship Certificate
TARS Technologies
AI / ML Internship — Completion Letter
Kohli AI
ML Engineer Internship — Offer Letter
Kohli AI
Data Science Master — Virtual Internship
EduSkills
EduSkills Virtual Internship
EduSkills
EduSkills Virtual Internship
EduSkills
Let's Work Together.
Open to AI/ML engineering roles, internships, and ambitious collaborations. Have an idea or a team that needs an extra brain on the model? Send a message and let's build something remarkable together.