2002
Flash and multimedia
Delhi studios
2006
Frontend architecture
Publicis Sapient
2014
Component platforms
NatWest Group
2022
Micro frontends at scale
Advisor360°
Now
AI agents and harnesses
Advisor360° and GitHub
Every era of the web, built hands-on. The frontend depth is what keeps the agent work grounded in shipping software.
What I'm building
Agents that ship real code
The hard part of AI engineering is everything around the model: what the agent may touch, who checks its work, and what proof it brings back.
How work moves through my agents
- Taska ticket, an upgrade, a repeatable chore
- Agent in a harnessskills · hooks · MCP tools
- Subagent panelindependent reviewersauditcross-checkverify
- EvidencePlaywright runs · evals
- Human reviewsees only verified findings
Harnesses and tooling
Skills, plugins, hooks and MCP integrations that turn repeatable engineering work into agent workflows someone can review.
Multi-agent orchestration
Subagents that audit and cross-check each other, so a finding is verified before a person spends time on it.
Model routing and governance
Open-weight models through OpenRouter, with zero data retention, cost tracked per session, and control over which region serves each call.
Agent-driven modernization
Agents carrying a large Nx monorepo through a major Angular upgrade, with Playwright checks and evidence in hand before release sign-off.
Experience
Twenty-four years, three long chapters
Dec 2022 – now 3 yrs 11 mos
Principal Software Engineer · Advisor360°
- Own the architecture of the Nx monorepo and Module Federation platform that Advisor360°'s WealthTech applications run on.
- Running the platform's Angular major-version upgrade with agents doing the migration and Playwright evidence gating every release.
- Build agent tooling for the engineering org: skills, plugins, MCP integrations and review panels of subagents.
- Leading a proof of concept for open-weight models on OpenRouter, with zero data retention, regional routing and cost measured per session.
- Coach engineers on working well with AI agents.
Mar 2014 – May 2022 8 yrs 3 mos
Principal Engineer · NatWest Group
- Built WebSDK, a React component library that cut web app development time by 30%.
- Built the bank's in-house deployment and release tool on React, microservices, AWS and PCF.
- Moved monolithic applications to single-page apps backed by microservices.
- Trained 20+ backend developers into frontend engineers.
Sep 2006 – Feb 2014 7 yrs 6 mos
Manager, Interactive Development · Publicis Sapient
- Designed frontend architecture for Merrill Lynch, TD Bank, Citigroup, ESPN, Coca-Cola and Vodafone.
- Wrote the JavaScript boilerplates that sped up development by 20%.
- Ran hiring, mentoring, code review and training for a 40+ person interactive studio.
2002 – 2006
Frontend Developer at R Systems · UI Designer at Brightpoint India · Web Architect at Innovative Software · Multimedia Designer at Angelus InfoTech. Flash, Director, ASP and hand-built HTML.
Delivered for
- Merrill Lynch
- TD Bank
- Citigroup
- Coca-Cola
- ESPN
- Vodafone
- JCPenney
- Michelin
- NCL
- Moët Hennessy
- Diesel
Toolbox
The stack I work in
Each layer is what the one above it stands on. The agent work only holds up because the platform work underneath it is solid.
AI engineering
what I build now
- Coding agents
- Agent harnesses
- MCP
- Subagents
- Evals
- Guardrails
- Model routing
Product platforms
what the agents work on
- Angular
- Nx
- Module Federation
- RxJS
- React
- Playwright
Foundations
under everything
- TypeScript
- JavaScript
- Node.js
- Docker
- AWS
- GitHub Actions
Let's talk.
Pick a topic and your email opens with the subject filled in, or just write.
