Open to AI/ML & Python Developer roles

Engineering Intelligence: Production AI Systems & High-Performance Python Solutions

AI/ML Engineer & Python Developer building production GenAI systems — RAG pipelines, agentic workflows, and document intelligence — with measurable impact: 70% less manual data entry, 50% lower API latency.

Explore AI Projects
Harsh Nanglia
Harsh Nanglia
rag_pipeline.py
70%
less manual data entry
50%
API latency reduction
100+
docs in extraction pipeline
15+
bugs caught pre-release
Featured Work

Systems I've shipped & built

From production EMS dispatch backends to self-directed RAG research — here's the work.

Production · EMS Dispatch

ETHAN — Emergency TeleHealth & Navigation

Real-time EMS dispatch backend for U.S. emergency services — owned API contracts, database schemas, and deployment end-to-end. Cut API latency 50% via SQL Server re-indexing and parameterised query batching. Integrated Azure Maps for live ambulance tracking and route optimisation across concurrent incidents.

FastAPIAzureSQL ServerAzure Maps
Production · Document Intelligence

IDE — Intelligent Document Extraction

Multi-modal extraction pipeline across 100+ heterogeneous PDFs and Excels, cutting per-document processing time ~70%. Composed Azure OpenAI, Document Intelligence, Tesseract and OpenCV under a strict JSON schema, validated with Plotly bounding-box overlays.

Azure OpenAITesseractOpenCVPlotly
Personal R&D · RAG System

DocuMind — RAG-Powered Document Q&A

Production-grade multi-document RAG pipeline using LLaMA 3, LangChain, ChromaDB, and FastAPI with hybrid BM25 + semantic retrieval, cross-encoder re-ranking, and citation verification. Features query reformulation, auto-summarization, conversational memory, and a retrieval explainability panel. Evaluated using RAGAS framework

LLaMA 3ChromaDBLangChainRAGASHuggingFace
Platform · Multi-Tool GenAI

KAIROS AI — Multi-Tool Generative Platform

Unified GenAI platform combining conversational AI, multi-PDF Q&A, voice assistance, and language utilities. Chat-with-multiple-PDFs module backed by OpenAI embeddings enabling natural-language exploration of large document corpora with follow-up queries.

OpenAI EmbeddingsJavaScriptREST API
Full-Stack AI · Safety Architecture

MindfulChat — AI-Powered Mental Wellness Companion

Responsible AI engineering platform with a two-layer safety system — combining keyword and semantic similarity detection (sentence-transformers + ChromaDB) that bypasses the LLM for immediate crisis routing. Features long-term vector memory, RAG-grounded coping strategies, AI journaling with emotion analysis, and mood-based TMDB recommendations.

React FastAPI Groq (Llama 3) ChromaDB Sentence-Transformers TMDB API Vercel + Render
Automation · Content Pipeline

AI News Video Generation Pipeline

End-to-end pipeline converting trending topics into short-form video assets via Gemini (script) and Imagen (visuals), using structured prompt templates and NLP post-processing for tone/shot-list consistency — session state persisted in Firebase Firestore.

GeminiImagenFirestoreStreamlit
Toolkit

Enhanced Skills Matrix

Filter by category to explore technical proficiencies across GenAI, Machine Learning, Cloud, and Engineering.

All Skills
GenAI & LLMs
ML & Deep Learning
Cloud & MLOps
Python & CV/OCR
Data & Databases
Journey

Work Experience

Nov 2025 — Present
Independent AI Engineering & Research
Personal R&D
  • Building DocuMind, a production-grade RAG system featuring hybrid search and automated reranking pipelines.
  • Deepening practical expertise in agentic AI frameworks, state-machine agent graphs (LangGraph), and LLM fine-tuning techniques.
Apr 2025 — Oct 2025
Python Developer — AI & Automation
Digital Glyde · Full-time, Onsite
  • Designed the FastAPI + Azure backend for ETHAN, a real-time U.S. EMS dispatch system — owning API contracts, schemas, and deployment.
  • Cut API latency 50% by re-indexing hot SQL Server tables and rewriting N+1 queries.
  • Built a multi-modal document extraction pipeline (IDE) across 100+ PDFs/Excels, cutting processing time ~70%.
Jul 2024 — Aug 2024
AI & Cloud Computing Intern
AICTE Edunet Foundation · Remote, India
  • Built a production-deployed conversational AI chatbot using IBM Watson Assistant with intent classification and entity recognition.
  • Architected scalable cloud infrastructure and deployment pipelines on IBM Cloud infrastructure.
Jun 2024 — Jul 2024
Data Analyst Intern
DATAPLAY · Jaipur (Remote)
  • Designed interactive Power BI dashboards, increasing decision accuracy by 35%.
  • Led a cross-functional team of four analysts under tight delivery deadlines.
Impact & Leadership

Hackathon Success & Innovation

National-level recognitions and social impact engineering projects.

National Level · Competition

Smart India Hackathon Finalist

Two-time finalist demonstrating exceptional problem-solving and innovation capabilities in national-level competitions under high-pressure environments.

Problem SolvingSystem ArchitectureNational Finalist
Assistive Tech · Social Impact

Tongue Driving System (TDS)

Led development of assistive technology for individuals with locomotor disabilities, showcasing technical leadership, hardware/software integration, and real-world social impact.

Assistive TechEmbedded ControlsSocial Impact
Government Mobility · Smart Transportation

Track-VAHAN

Engineered a smart transportation system for the Himachal Pradesh government featuring real-time vehicle tracking, route optimization, and automated accident detection algorithms.

Real-time TrackingRoute OptimizationAccident DetectionGovt Project
Background

Education

Jun 2021 — Jun 2025
B.Tech, Computer Science & Engineering
Maharaja Agrasen Institute of Technology, Baddi

Academic Performance: CGPA 8.50

Get in touch

Let's build something

Open to opportunities & collaborations

Whether it's a full-time AI/ML role, a RAG system that needs building, or a quick chat about agentic workflows — reach out.

H
Harsh AI
Ask about my work