I am a Software Engineer specializing in AI and industrial automation.

> about.me

/bin/bash

user@host:~$ cat user_profile.txt

A photo of Harsh Kumar

> Name: Harsh Kumar

> Status: Looking for Opportunity

> Location: New Delhi, India

terminal

Welcome to Harsh Kumar's Interactive Portfolio Terminal (v1.0.0)

Type help to view all available commands.

guest@harsh: ~ $

> git log --experience

commit d4a2c1a

Full-Stack Engineer

GLBE LLP

Active

Mar 2025 — June 2026

  • → Architected and deployed a secure portal that automated vendor onboarding and invoice processing, reducing manual workflow time by 70%.
  • → Designed RESTful APIs with JWT authentication and RBAC, enforcing zero-trust security and strict data isolation across workflows.
  • → Built a real-time financial dashboard tracking invoice statuses (settled payments, outstanding balances, held disbursements), enabling stakeholders to monitor cash flow at a glance.
RESTful APIs JWT Auth RBAC Zero-Trust Financial Dashboard Workflow Automation
commit b8e9f3d

Frontend Developer Intern

ApexIQ AI

Aug 2024 — Feb 2025

cat proof.log
  • → Architected and shipped reusable React component library, enabling faster feature development across the team.
  • → Reduced page load latency by 25% via code splitting, asset minification, and lazy loading.
  • → Improved application stability through systematic debugging and state management refactoring, reducing production issues.
React.js Component Library Code Splitting Lazy Loading State Management Optimization
commit h7e1f3h

Web Developer

Zen Legal

June 2024 — Aug 2024

cat proof.log
  • → Desgin and deployed a responsive website from scratch, establishing the platform's initial digital footprint.
  • → Engineered core frontend features and user flows, transforming project requirements into a functional, production-ready site.
  • → Collaborated with cross-functional teams to debug, refine, and deliver all assigned development milestones within strict timelines.
HTML CSS3 JavaScript Tailwind CSS

> ls -a ./projects

app/overlay-meeting-copilot.ts
Live App
Overlay AI Real-Time Meeting Co-Pilot Preview

Overlay AI: Meeting Co-Pilot

Local BYOK Engine

Real-time desktop meeting assistant built with Electron.js and BYOK LLM architecture. Captures local audio transcripts, generates automatic MoM (Minutes of Meeting), and provides contextual on-call guidance.

Electron.js React.js BYOK Gemini API Claude Local Relayer
src/docmind-agentic-rag.py
Featured
DocMind Agentic RAG System Preview

DocMind: Agentic RAG System

35% Latency Saved

Enterprise Agentic RAG system built with LangGraph for multi-step reasoning and dynamic retrieval. Integrated ChromaDB semantic search and Redis caching to minimize LLM token usage.

LangGraph Gemini React.js FastAPI ChromaDB Redis
src/shoppy-assist.py
Featured
Shoppy Assist AI Shopping Assistant Preview

Shoppy Assist: AI Assistant

Ollama + Scraping

AI shopping assistant combining Selenium & BeautifulSoup real-time e-commerce scraping with a locally hosted Ollama (Llama 3.1) LLM for personalized price analysis in Streamlit.

Ollama Llama 3.1 LangChain Streamlit Selenium Python
reports/sales-bi.pbix
// analytics
Myntra Sales Performance Dashboard Preview

Myntra: Sales Performance

$1.9M Sales Data

Interactive Business Intelligence dashboard modeling $1.9M in e-commerce transaction data to uncover city-wise revenue drivers, customer demographic age profiles, and category trends.

Power BI Excel Power Query Data Modeling DAX
bin/eagle-eye.py
// vision
EagleEye Smart Surveillance Preview

EagleEye: Smart Surveillance

YOLOv3 + ALPR

AI surveillance system integrating real-time Weapon Detection, Fall Detection, Crowd Counting, and Automatic License Plate Recognition (ALPR) via OpenCV & YOLOv3.

Python OpenCV YOLOv3 ALPR Fall Detection Crowd Count
bin/autoagent-cli.py
// agent
AutoAgent CLI Agent Framework Preview

AutoAgent: CLI Framework

Docker Sandboxed

Autonomous multi-agent execution framework using LangGraph to compile objectives into task DAGs, execute Python scripts inside sandboxed Docker containers, and self-correct on error.

LangGraph Python Docker OpenAI API Shell

> git stats --global

-- // repositories
-- // total stars
-- // followers
642 // commits (YTD)

// note: metrics reflect public repositories only. private repository data is not visible.

// contribution_calendar.sys
Less More

> ls /bin/skills

// Languages & Web

Python
JavaScript
HTML5
CSS3
React.js
Flask
FastAPI
Tailwind

// Databases & Infrastructure

PostgreSQL
Redis
MySQL
Docker
Git/GitHub

// AI & Machine Learning

🦜
LangChain
OpenCV
🦙
Llama 3.1
✨
Gemini
🤖
Claude / GPT

// GenAI Methodologies

🔍
RAG
✍️
Prompt Eng.
🧠
Context Mgmt.
📦
Output Struct.
⚡
Token Opt.

> contact --send

Have an idea? Let's turn it into a real-world solution.