VVarun NidhiApplied AI & LLM Solutions Architect
Helping people make better calls under uncertainty

Varun Nidhi — I architect AI solutions
that turn messy operations into decisions teams trust.

I architect AI solutions for the hard end of real operations — anywhere the data is incomplete, the conditions are unforgiving, and a wrong call is expensive. I take it the whole way: discovery, architecture, model selection, integration, evaluation, rollout — and keep the boardroom and the engineers speaking the same language throughout.

Explore the toolsSee the case studies
Varun NidhiVarun Nidhi
01Financial Tools

Tools for making sense of markets — the same instinct, pointed at price.

TradingView Indicators

A growing library of Pine Script indicators for reading markets under uncertainty — published openly on TradingView for anyone to add to their charts.

Explore
Open Options Trader

A free, educational web app for exploring Indian F&O contracts — price options with Black-Scholes, browse expiry calendars, and build multi-leg strategies with payoff diagrams. Theoretical model prices, not live market data.

Open Options TraderExplore
Pigeon Station

A self-hosted station that turns financial news — RSS feeds and Telegram channels — into structured, filterable trading signals, with an LLM doing the reading so you're left with the call, not the noise.

Explore
02Creative Tools

Tools for making things — the same instinct, pointed at creative work.

Creator Lab

A self-contained creative-learning launcher — a Steam-style catalog of 64 curated tools and services across 15 creative categories, plus 8 guided learning paths that chain them into step-by-step projects like making your first short film or building a 2D game.

Creator LabMake
03Open Source

A few things I've open-sourced — free for anyone to pick up and run.

LLM-on-Web

An AI chat app that runs language models entirely in your browser — no server, no sign-up — with document Q&A built in.

View repoLive demo ↗
Prompt2Powerpoint

Turn a plain-language prompt into a finished PowerPoint deck, working with either a local model or a cloud API.

View repoLive demo ↗
QuickTag-Images

Point it at a folder of images and a local AI gives every file a sensible name and tags — organisation without the busywork.

View repoLive demo ↗
Browse github.com/van007
04Things I've Built

Products built to survive contact with the real world.

Pipeline Operations Console

Oil & Gas · LLM

An AI console that helps teams reason over pipeline inspection and operations data — and turn it into action in one click. The LLM does the work; no one writes a single prompt.

Pipeline Operations Console

Oil & Gas · LLM

An AI console that helps teams reason over pipeline inspection and operations data — and turn it into action in one click. The LLM does the work; no one writes a single prompt.

Problem

Pipeline inspection and operations data shows up in volume and in pieces — survey files, inspection logs, spreadsheets, SCADA, and more. Making sense of it is slow, manual work, and the answer is only ever as good as whoever had the patience to read every file.

Approach

I built a console that reads across inspection records, operations data, and GIS data and makes sense of it together — so a team can ask a plain-language question, get an answer grounded in their own records, and turn it into action in one click. Evaluation loops track how good the answers are and how fast people reach a decision as they use it.

Impact

  • One-click reports replace hours and days of manual assembly
  • Teams reason over inspection and operations data by asking, not hunting through files
  • Adoption and answer quality tracked with built-in evaluation loops

Crude Trade Copilot

Oil & Gas · LLM

A copilot for crude trading — read an assay (its specifications), work out the blend, predict the yield, all from a plain-language prompt.

Crude Trade Copilot

Oil & Gas · LLM

A copilot for crude trading — read an assay (its specifications), work out the blend, predict the yield, all from a plain-language prompt.

Problem

Buying crude oil starts with its assay — a detailed breakdown of what the crude contains. From there a trader has to work out how to blend it and predict what it will yield once refined. It's specialist work, done under time pressure, and a wrong call moves real money.

Approach

I built a copilot that reads the assay, works out the blend, predicts the yield, and writes a clear summary ready for a decision — all from one plain-language request. Work that used to take a specialist hours now happens in a single step.

Impact

  • Days of work assessing the right crude, now done in minutes
  • Consistent results every run — no prompt-wrangling, no guesswork
  • Fast enough to move on a high-margin spot crude before the window closes

Agentic Reporting Platform

Agentic AI · LLM

An agentic platform that turns weeks of client reporting into hours — AI agents read a team's own data and draft exec-ready output, leaving a person to review, not assemble.

Agentic Reporting Platform

Agentic AI · LLM

An agentic platform that turns weeks of client reporting into hours — AI agents read a team's own data and draft exec-ready output, leaving a person to review, not assemble.

Problem

Client reporting is slow, repetitive work. Pulling the numbers, assembling the narrative, formatting it for an executive audience — it can eat weeks, and the people doing it are the same specialists you'd rather have solving the underlying problem.

Approach

I led an agentic platform where AI agents read across a team's own data, work out what the report needs to say, and draft it exec-ready, end to end. Work that used to take weeks of assembly now takes hours, with a person in the loop to review and sign off rather than build from scratch.

Impact

  • Client reports in hours that used to take weeks
  • Specialists review and sign off instead of assembling from scratch
  • Exec-ready output drafted from a team's own data, end to end

Field Mobility Apps

Field Operations · Mobile

Mobile apps that carry multimodal field data — photos, notes, audio, video — back to a central server, then return the analysis as clear action points.

Field Mobility Apps

Field Operations · Mobile

Mobile apps that carry multimodal field data — photos, notes, audio, video — back to a central server, then return the analysis as clear action points.

Problem

Field teams see problems first, and they capture them however they can — a photo, a scribbled note, a voice memo, a quick video. But that's usually where it stops: the richest signal about what's happening on the ground sits trapped on a dozen phones, never reaching the people who could act on it.

Approach

I built mobile apps that let field professionals capture multimodal data — images, text, audio, video — and send it straight to a central server for analysis. From there it flows two ways: management gets the picture they need, and the field team gets specific action points back. The loop from observation to decision to action finally closes.

Impact

  • Field observations reach a central analysis instead of staying stuck on phones
  • Management sees what's happening on the ground without waiting for a report
  • Field teams get concrete action points back, not just an acknowledgement

Smart Education Apps

Education · Mobile + AR

Mobile apps that connect printed textbooks to digital learning — point a phone at a page to unlock videos, quizzes, and augmented content. One of the first of its kind in the country.

Smart Education Apps

Education · Mobile + AR

Mobile apps that connect printed textbooks to digital learning — point a phone at a page to unlock videos, quizzes, and augmented content. One of the first of its kind in the country.

Problem

A printed textbook is fixed the day it goes to press. Everything that makes a subject click — a video, a quick quiz that checks you actually understood, a model you can turn around in 3D — lives in a different world entirely, on a screen the book has no way to reach.

Approach

I built mobile apps that turn a printed page into a doorway. Point a phone at the book and it brings up the video, the quiz, the augmented content tied to exactly that lesson. It was one of the first projects of its kind in the country — printed and digital learning finally pointing at the same thing.

Impact

  • A printed page becomes a launch point for video, quizzes, and augmented content
  • Students keep the book they already have — the digital layer meets them there
  • One of the first printed-to-digital learning projects in the country

Want to hear more?

These are a few of the things I've built, with more on the way. If you've got a messy problem of your own — or just want the longer story behind one of these — I'd like to hear from you.

Get in touch
05Approach

How I take AI from idea to something people rely on.

01
Discovery
Find the real problem worth solving — and whether AI is even the right tool for it.
02
Architecture
Design how the whole solution holds together, and how it feels to use day to day.
03
Model selection
Pick what actually fits the job — local, on-prem, or cloud — not what's fashionable.
04
Integration
Wire it into the systems and data people already work in, so it lands in real use.
05
Evaluation
Build the frameworks that track answer quality and time-to-decision as people use it.
06
Rollout
See it through to daily, trusted use — and keep refining once it's live.
06About

What separates an AI demo from a system people trust.

01

Beyond the demo

Most AI projects stall at the demo. I care about the part that comes after — turning a promising model into a system people trust.

Beyond the demo

Most AI projects stall at the demo. I care about the part that comes after — the unglamorous work of turning a promising model into a system people open every morning and trust. So I architect the whole path: finding the real problem, designing how the solution holds together and how it feels to use, choosing the model, wiring it into the systems people already work in, building the evaluation that proves it works, and seeing it through rollout.

02

What I'm building now

AI for live industrial operations — turning volume and contradiction into a call someone can stand behind.

What I'm building now

Right now that's a set of AI solutions for live industrial operations: a console that lets a team reason over scattered inspection data instead of hunting through every file, a copilot that turns a specialist's workup into a single plain-language prompt, and a dual-path setup that runs inference on the device itself, so sensitive data never has to leave the building. The industry behind them matters less than the hard part they share — turning volume and contradiction into a call someone can stand behind.

03

Fifteen years of it

Digital twins, defense, field mobility, education, finance, connected devices — different domains, one thread running through all of them.

Fifteen years of it

The years before that went into software for other people doing consequential work, across industries that look nothing alike. A decade of it was digital twins — working simulations of real pipelines, used to keep them safe and running. The rest ranged wide: big-data engines reading over a million data points a second and analytics across billions of records for defense, mobility apps built for field professionals, education apps that bring printed books to life on a screen, finance apps, and IoT backends for connected devices. Different domains, different users — but the thread is the same, and it's the part I genuinely enjoy: the data shows up messy and contradictory, and what matters is whether someone can look at what the system says and make a confident call. Architecting the path to that moment is the work I find most satisfying.

Since 2010
Fifteen years in industrial tech — from wireless to AI
Proven at scale
Engines that read 1.3M data points a second, analytics across billions of records, and reporting that went from weeks to hours
Across industries
Pipeline safety, defense, field mobility, education, finance, connected devices, and big data
07Contact

Have a problem worth solving? Let's talk.

LinkedIn

in/varunnidhi

↗

GitHub

@van007

↗

Based in India