Harsh Vardhan Singh Profile

Hi, I'm Harsh Vardhan Singh

frontend • backend • ai & agentic systems

I'm a software engineer and builder who enjoys taking ideas from a blank screen to production. I care about understanding how things work under the hood. I'm happiest when I'm learning something new, building something challenging, and shipping it to real users.

Harsh Paper Disco Visual
Recent Activity
Repos
Harsh Vardhan Singh Official GitHub Contribution Graph
[ Places I've Made an Impact ]
Full Resume (PDF)

UnderAI

Full-Time
Software Developer
Oct 2025 — Present
Remote / Delhi NCR

Leading AI systems development, agentic workflows, and document intelligence pipelines for enterprise risk & RFQ automation.

Built a multi-agent risk intelligence system performing live industry research, document parsing, and hazard analysis—enabling clients to upsell relevant coverages by 50%+ and accelerating deal closure.
Architected an AI document analysis pipeline ingesting and correlating historical policy data, exposure documents, and inspection reports to extract structured business-critical information.
Built a multi-stage agent orchestration workflow using LangChain and LangGraph for document understanding, rule-based reasoning, and decision synthesis.
Designed an AI-powered RFQ automation system with an interactive spreadsheet-style playground, cutting manual RFQ creation time from ~140 hours to ~15 minutes.
LangChainLangGraphMulti-Agent AINext.jsTypeScriptPythonDocument IntelligenceRFQ Automation

Freelance & Contract Work

Contract / Freelance
Full-Stack & AI Systems Developer
2024 — Present
Remote

Partnered with independent clients and early-stage startups to deliver high-converting web applications, custom AI integrations, and full-stack platforms.

Next.jsReactNode.jsExpressC++PythonTailwind CSSPostgreSQLVertex AI
[ Tech Stack ]The tools behind my builds
Frontend & Frameworks[6]
Next.jsReactTypeScriptTailwind CSSFramer MotionRedux
AI & Systems[6]
LangChainLangGraphGemini VisionVertex AIPythonC++
Backend & Databases[6]
Node.jsExpress.jsPostgreSQLMongoDBPrismaDocker
[ Things I've Built ]
Vitch.AI — AI Agentic Video Editor
FEATURED AI

Vitch.AI — AI Agentic Video Editor

Browser-based, AI-assisted nonlinear video editor combining multi-track editing, WebGPU preview compositor, FFmpeg export, and an AI Creative Director.

Next.jsTypeScriptAI AgentsWebGPUPythonFFmpeg
Ocular — AI Document Verification Engine
AI ARCHITECTURE

Ocular — AI Document Verification Engine

Extract text from PDFs, send to LLMs, and automatically highlight AI responses back to their exact position in the PDF with ZERO 2nd LLM call latency.

TypeScriptPythonPDF EngineLLMsZk-Proof Attribution
Timeloom — Prospect Outreach SaaS
SAAS PLATFORM

Timeloom — Prospect Outreach SaaS

Productivity & outreach automation tool allowing users to connect across tools and streamline multi-channel prospect messaging.

Next.jsTypeScriptPostgreSQLTailwind CSS
Shrenik Escapes
LUXURY TRAVEL

Shrenik Escapes

Luxury resort & experiential travel booking platform engineered with Next.js, smooth animations, and high-converting visual storytelling.

Next.jsTypeScriptTailwind CSSFramer Motion
Fuji Yatra
TRAVEL PLATFORM

Fuji Yatra

Trekking & pilgrimage travel portal featuring interactive itinerary exploration, booking workflows, and optimized fast-rendering dynamic routes.

Next.jsReactTypeScriptTailwind CSS
Snatchedit — AI Virtual Try-On
AI & VISION

Snatchedit — AI Virtual Try-On

Paste any fashion product link and visualize yourself wearing it using Gemini Vision & Vertex AI computer vision pipelines.

TypeScriptGemini AIVertex AINext.js
[ Technical Writing & AI Research ]
2 Publications
cat ~/writing/pdf_attribution.md
AI ATTRIBUTION
Medium Publication
6 min read

“Your AI read the document, can it prove it?”

Zero-Latency PDF Highlight Attribution in Production RAG

How to eliminate LLM citation hallucinations by extracting exact PDF bounding box coordinates in real-time without secondary model overhead.

0 Secondary LLM OverheadBounding Box VerificationProduction-Grade PDF Parsing
git clone github.com/BerrySeriousCoder/eval
EVALS & OBSERVABILITY
GitHub Repository
Open Source

Evaluating, Monitoring & Deploying Production AI Systems

Open Technical Benchmarks & Evals Framework

In-depth technical research, code patterns, and latency/hallucination tracking guidelines for monitoring enterprise AI deployments.

LLM Evaluation BenchmarksLatency & Token Cost TracingHallucination Detection Guardrails
[ GET IN TOUCH ]

Let's build something exceptional together.

I am currently open to full-time engineering roles, freelance projects, and AI product collaborations.

harsh100xdev.work@gmail.com