ASHISH AHUJA · ENGINEERING LEADER · MUMBAI

I lead engineering teams at work. In 2026 I also built 29 apps end to end with AI.

11+ years shipping enterprise software for multi-tenant SaaS and Swiss banking. Every product on this page was planned, designed, built, tested, deployed and is still run by me, with Claude Code agents doing the typing. Three are live demos you can use right now, with sample data.

TYPICAL DAY12:00 ai-office stand-up done · 5 tasks picked
29
apps built with AI (23 web, 6 Android)
740+
pull requests merged
34
AI agents running one office
15
engineers led at my day job
Use them, don't just read about them

Live demos

Real apps with sample data only. Add entries, change things, try to break them. Each demo resets every night.

AI OFFICE · MULTI-AGENT SYSTEM

A company of 34 AI agents

A Chief of Staff, CTO, QA Manager, Release Manager, Scrum Master, sales and marketing agents run a working day on their own.

  • 12:00 stand-up: the Scrum Master picks the day's tasks within a set pace
  • Owners work one task at a time; QA checks anything in review first
  • Evening briefing email; pauses itself near the usage limit and resumes
  • Chat with any leader from a PIN-protected page, by text or voice
Claude Code (headless)Node.jsAgent rolesSchedulerEmail
SAMPLE DATA
PERSONAL FINANCE · WEB APP

Daily Expenses

Money the way an Indian household actually runs it: many bank accounts, many credit cards, EMIs on cards, and money lent to friends.

  • Accounts, cards, monthly statements and bills with part payments
  • EMI schedules with interest and GST on interest, linked to card statements
  • Money owed both ways, multi-currency accounts with a live rate
  • Hinglish voice entry, and a Google Sheets sync with a check that flags any mismatch
ASP.NET Core 8Angular 19SQL ServerGoogle Apps ScriptSpeech-to-text
SAMPLE DATA
SMALL MANUFACTURING · ERP + KIOSK

Factory Ledger

A small factory's books, from the sales bill to the worker's salary. It replaced the owner's Excel workbooks.

  • Sales with many items, purchases, party ledgers and payments
  • Bills sent on WhatsApp as a picture card, through the Meta Cloud API
  • Product photos and a public catalogue for buyers
  • Gate kiosk: workers check in by QR, face or PIN; monthly salary from attendance
ASP.NET Core 8AngularSQL ServerWhatsApp Cloud APIface-api.js
SAMPLE DATA
Everything shipped in 2026

All products

Each one started as a real problem from a real person, went through a clickable mock first, and is live or in daily use.

Tech
AI-driven development

How I build with AI

The same seven steps for every change, small or large. AI writes most of the code; the decisions, reviews and releases stay with me.

  1. Hear the problemThe owner explains it in plain Hindi or English. One question at a time, no guessing their numbers.
  2. Clickable mockNothing is built before the owner taps through a mock and says OK.
  3. Build on a branchClaude Code agents write the code on a feature branch, inside written house rules.
  4. TestAPI checks and headless-Chrome flows at phone and desktop widths, against a copy of the database.
  5. Pull requestsFeature to develop to main in Azure DevOps. Only main can deploy.
  6. Pipeline deployBackup first, deploy from the pipeline, then check the live site.
  7. Watch and learnA watchdog, nightly backups, and a knowledge file of every decision.
Azure DevOps · CI/CD

Every change ships the same way

A feature branch, two reviewed pull requests, then a pipeline that backs up, deploys and checks the live site. Only main can reach production.

Branches and pull requests sample run

maindevelopfeature/voice-entry PR · reviewed · merged PR → main starts deploy
  • feature/…one change, tested against a copy of the database
  • developreviewed changes wait here together
  • mainwhat is live; a merge here starts the deploy

DailyExpenses-Deploy · main Succeeded

  1. Check out main
  2. Restore packages
  3. Build API (.NET 8)
  4. Build web (Angular)
  5. Back up the database
  6. Deploy to IIS
  7. Check the live site
745pull requests
26CI/CD pipelines
29git repositories
14Azure DevOps projects

Counted from the Azure DevOps API on 15 Nov 2026. First pull request: 31 May 2026.

Infrastructure I set up and run

One server runs it all

I moved 14 production apps from Docker and PostgreSQL on Linux to this Windows server, one app at a time.

UsersPhones and laptops: owners, workers, customers
CloudflareDNS and a tunnel, so the server opens no inbound ports
Windows Server 2025 VPS
IIS
23 live sites
APIs
ASP.NET Core 8, Node.js
Web
Angular apps
Data
SQL Server, a database per app
Azure DevOpsRepos, pull requests, and a pipeline per app on a self-hosted build agent
Claude CodeDevelopment sessions and the scheduled AI office
WatchdogSites, services, memory, disk and backups; alert emails and a status page
The day job

Career

Engineering leadership in multi-tenant SaaS and Swiss banking. Certified Scrum Master.

Technical Lead, ZI SystechLeads 15 engineers on a multi-tenant survey and analytics SaaS: architecture, code-review gates, releases on Azure DevOps.
Developer / Team Lead, TCSLed an Agile team of 10; built a secure defect and vulnerability tracking platform from scratch.
Technology Analyst, InfosysBanking compliance and approval systems; 15 months on-site in Zurich with Swiss banking clients.