Customer success · Enterprise solutions · Voice and conversational AI
I put AI agents into production inside large enterprises.
Banks, insurers, carmakers, manufacturers. I find the use case worth building, write the integration spec, build the demo, and stay until it runs at scale. Then I take the numbers to the C-suite.
About
The model is usually the easy part.
What decides whether an AI agent survives in a bank or an insurer is everything around it. The dialler. The CRM write-back. The ninety ways a human agent closes a call. The compliance team that has to sign off. That's where I spend my time.
Right now I lead customer success and enterprise solutions at ConvoZen.ai, NoBroker's voice AI platform. I own a book of 20+ enterprise accounts across BFSI, insurance, auto, real estate and manufacturing, and lead a small team that takes them from POC to production.
I got here the long way. I opened a B2B business from zero at OYO, a year out of engineering school. Did my MBA at Schulich in Toronto. Ran two fulfilment centres for Amazon in British Columbia with 700+ people. Then went back to building, first at Gupshup and now at ConvoZen.
I'm hands-on. Python and SQL most days, and Claude Code is open on my laptop from morning to night. Most of the demos, specs and analyses on this page I built myself.
Deployments
Where it's running
A selection from the last three years. Each card names the team the agent works for and the outcome I can share publicly.
HDFC Life
Renewal collections calling plus an inbound service agent, in 11 Indian languages.
For: Collections and Persistency, Customer Service
96x monthly interactions in one quarter
Read the case › BankingJana Small Finance Bank
EMI reminder and collections voice agents at scale, plus a dialling analysis that changed how the bank calls.
For: Collections
3.36M calls a month, 1.5M customers reached
Read the case › Solar energyWaaree Energies
Four voice agents that take a solar enquiry from CRM lead to paid order, with no human on the line.
For: Sales, Channel Partners
41 post-call extractors feeding the CRM
Read the case › AutomotiveMaruti Suzuki
"Maya", an inbound voice receptionist for dealerships that answers model questions, handles competitor talk and books test drives, carrying on over WhatsApp.
For: Sales, Dealer network
Accelerator POC presented to Maruti leadership
Read the case › Real estateCasagrand
A property discovery and lead-qualification agent that knows every live project, unit and price across Chennai.
For: Sales
Award Voice AI Pioneers
Read the case ›Karur Vysya Bank
Business case for pre-due collections and call-quality monitoring with voice AI.
For: Collections, Quality
~47% lower cost per minute than human calling
Also: a bilingual English and Hindi sales agent for Niyo's premium travel card. Earlier, at Gupshup: voice AI and CPaaS across enterprise accounts and high-growth startups, taking the portfolio to $12M+ ARR.
On the floor
Voice AI Pioneers
Accepting the Voice AI Pioneers award for our real estate discovery and lead-qualification agent, at our event with EkStep.
Maruti Suzuki Accelerator, presenting the POC · HDFC Life AI Day, running the live demo · ConvoZen launch day
Things I've built
I prototype before I propose.
Internal tools that my team and customers use every week. All built with agentic coding tools, mostly Claude Code.
Analyst agent over 68M calls
Ask a question in plain English, get the answer as a table. The agent knows the schema, writes the SQL, runs it against Metabase and formats the result. Its prompt carries the hard-won rules: which filters are mandatory, which timestamps are in epoch milliseconds, which fields lie.
Claude Code · Metabase API · Bash · Node · custom skills for transcript analysis
MBR and proposal generator
A Slack bot. Ask it for a monthly business review, an SOW or a proposal for a client, and it pulls live usage from Metabase, has Claude write the narrative, and posts a finished PDF, Word or PowerPoint file back in the thread.
Node · Slack Bolt · Claude API · Puppeteer · pptxgenjs
Delivery tracker
The system my team runs on. Every account from POC to onboarding to production, with subtasks, meeting notes, action items and a sales pipeline next to it.
Next.js 14 · TypeScript · Prisma · Postgres · NextAuth · Docker
ConSensus
An internal prediction market. Employees trade virtual coins on company outcomes (will this deal close this quarter?), and the price becomes a forecast leadership can read.
React · Vite · Express · Prisma · constant-product market maker
Path
- 2025 – nowConvoZen.ai (NoBroker)Associate Director, Customer Success and Enterprise Solutions · Bengaluru
20+ enterprise voice and conversational AI deployments. Lead a four-person team across 24 accounts.
- 2023 – 2025GupshupAssociate Director, GTM and Customer Success
Took a voice AI and CPaaS portfolio to $12M+ ARR. Moved GenAI workflows from pilot to production with enterprise sponsors.
- 2021 – 2023Amazon CanadaOperations Manager, L6 · Pathways · British Columbia
Ran two underperforming fulfilment centres. 700+ associates, 9 managers, seven promoted. $8M of cost out in five months.
- 2019 – 2020Schulich School of BusinessMBA · York University, Toronto
- 2015 – 2018OYO RoomsRegional Business Head · Chennai
Built the B2B vertical from zero to 400+ corporate accounts. Scaled supply from 10 to 500+ properties. Hired the team and wrote the playbook.
- 2014NIT TiruchirappalliB.Tech, Electrical and Electronics Engineering
Off the clock
IPL Fantasy
A fantasy cricket league for friends, with live scorecard scraping, a scoring engine and private leagues. Match-day operations run through Claude Code slash commands.
Next.js · Supabase · Vercel · see it live
A trip, engineered
Sixteen days across Vienna, Salzburg, Prague and Budapest. It became a generated print guidebook with its own map engine, plus an offline-first travel app for the four of us.
Python · Next.js PWA · Supabase · Leaflet