And the products and websites around them. Nine years building software in-house, most recently leading engineering at VMock. Now independent, working with founders and teams who need it done properly.
An agentic workflow that turns a client’s CSV into the configuration a resume product needs: setup from an hour to five minutes, customization from an hour to ten.
AI
Product
client setup, from 1 hour
5 min
editor customization, from 1 hour
10 min
2025
Employer branding page automation
Seventy per cent of the manual effort gone from employer branding pages, with an LLM workflow that researches and drafts them, and leaves the judgment to a person.
AI
Product
reduction in manual effort
70%
2022–2024
Employer platform revamp
Three new modules in front of more than five thousand employers, built by a team of four-plus without the three drifting into three different products.
Web
Product
employers on the platform
5,000+
developers led
4+
2026
Security questionnaires answered from company documents
A retrieval agent that answers enterprise security questionnaires from the company’s own documents, with every model call masked and every outbound request logged.
AI
Product
clients' security questionnaires and RFP responses I drafted by hand
35+
Show 2 more
2025
Orchestrator agent for an admin chatbot
Eight weeks to an agent in front of a client, answering admin questions against live platform data, with every run traceable afterwards.
AI
concept to client demo
8 weeks
2019–2021
Frontend build and bundle optimization
Builds 20% faster and the initial bundle 15% lighter on a large React codebase, working from bundle analysis rather than from the usual list of suggestions.
Web
SEO
build time
-20%
initial bundle size
-15%
PL 03
Services
100 px
Usually two to four weeks, architecture through to a deployed feature.
Put an AI feature in front of real users
Teams whose AI feature has to answer to a real customer, an auditor, or both.
You have a feature that needs a model behind it, and a demo is not the same as a thing customers use. I build the agent and the parts that make it accountable: what it can see, which tool ran, where a person signs off.
Agent architecture and orchestration (LangGraph, LangChain)
Streaming interfaces over WebSockets
Tool and API integration
Evaluation, tracing and diagnostics (Langfuse)
FastAPI services and deployment
Usually two to three weeks, design through to launch.
Ship a site that loads fast and stays fast
Founders and small teams who want to stop ringing a developer to change a price.
A site you can change yourself, that loads fast on a phone in a regional town and scores 100 for accessibility because it was built that way, not audited afterwards.
Design and front-end build (Next.js, React, TypeScript)
Design systems and reusable components
Performance work — Core Web Vitals, bundle size, build times
Accessibility to WCAG AA
Deployment and CI
Usually a four-week module, scoped before it starts.
Turn a vague idea into a shipped scope
Teams who know the problem, have a date, and need one person accountable from the scoping to the handover.
Someone has to turn “we need something like this” into a thing that ships, and then still make sense to the next person who opens it. Usually that means deciding what not to build.
Requirement refinement and scoping
Configuration-driven architecture for multi-client products
Cross-functional delivery with design and QA
Technical leadership and code review
Launch and handover
Usually a one-week audit, then fixes scoped from what it finds.
Make the site legible to search engines
Teams whose site is slow, badly rendered, or invisible to search.
The engineering side of search: rendering, speed, structured data and crawlability. The parts that are actually code, measured rather than promised.
Rendering strategy (SSR, SSG, ISR) for indexability
Core Web Vitals and page speed
Crawl, indexation and site structure
Structured data and metadata
Migrations planned to protect rankings
I kept no analytics for the SEO work I have shipped, so I make no traffic, ranking or revenue claims here. What I can show is the engineering: two client sites scoring 100 for Lighthouse SEO and accessibility, with structured data and four-locale hreflang. Judge the work by that, not by a number I cannot substantiate.
PL 04
How I work
100 px
Step 1
Scope
Before any code, a written scope that names the non-goals as plainly as the goals. Most projects do not fail on what was built. They fail on what nobody agreed to leave out.
You get: A scope document and a fixed estimate
Step 2
Build in the open
There is a deploy in the first week and something working to look at every week after. Not a slide showing what it will do, but a URL you can open, use, and forward to someone whose opinion you need.
You get: A staging URL from week one
Step 3
Harden
Error states, cost caps, evaluations and monitoring, before anyone calls it done. This is the part that separates a feature from a demo, and it is the part that gets cut when the build was not scoped to include it.
You get: A load and failure report
Step 4
Hand over
Documentation, a runbook and a recorded walkthrough, so the thing keeps working when I am no longer on it. The measure of a good handover is that nobody has to call me to change a value.
You get: A repository your team can actually run
What I turn down
Fixed-price work with undefined scope, and anything where the honest answer is that it does not need a model. If a cron job and three queries would do it, I will tell you that before you have paid me to find out.
PL 05
About
100 px
Gurugram, India
I have spent nine years building software that other people depend on, most of
it at VMock, where I led engineering on a career platform used by 20,000+
students and 5,000+ employers. I left in August 2026 to work independently.
The work I care about most sits where AI meets a real product. Not demos: an
orchestrator agent I built in eight weeks and put in front of a client, with
routing, memory, streaming and a human in the loop, because the interesting
problems only show up once something is actually in production.
Before the AI work I spent years on the unglamorous parts that make a product
good: build times, bundle size, the four hundredth UI bug, the design system
nobody notices when it works. That is still how I build.
These days I build for clients directly. A motel and a takeaway in regional New South Wales, both shipping in four languages, both scoring 100 for accessibility and SEO.
Small clients deserve the same engineering as large ones, and they rarely get it.
PL 06
Experience
100 px
Software EngineerSE
SSE
Engineering ManagerEM
20172019202120232026
Aug 2026 — Present (2 mo)
Independent Software Consultant · Freelance
Websites and AI features for clients directly, from architecture through to deployment.
Designed and shipped a four-language site for a boutique motel (alinemotel.com.au) in Next.js, TypeScript and Tailwind, with a Git-backed CMS and live room rates through a cached booking-engine proxy. It scores 100 for Lighthouse accessibility, best practices and SEO, held in CI by 110 end-to-end tests, and the owner updates all content without a developer.
Built a four-language restaurant site (apracurryhouse.com.au) in Next.js and Keystatic on AWS Amplify, with a weekly Playwright scraper that keeps the 102-dish delivery menu in sync and accessibility tests gating every page in light and dark. It scores 100 for Lighthouse accessibility, best practices and SEO, and the owners never update the menu by hand.
Aug 2021 — Aug 2026 (5 yr)
Engineering Manager · VMock
Led engineering across the platform: AI workflows, employer and candidate products, infrastructure and delivery.
Built an orchestrator agent for the admin chatbot in 8 weeks for a client demo, with LangGraph for context, tools, routing, memory, streaming and human-in-the-loop, Langfuse for diagnostics, FastAPI and ReactJS over WebSockets.
Automated content workflows with LangGraph, LangChain and FastAPI, auto-generating employer branding pages and cutting manual effort by 70%.
Led 4+ developers revamping the Employer platform (Job Author, BestMatch and ResumeBooks), serving over 5,000 employers.
Jan 2021 — Aug 2021 (7 mo)
Senior Software Engineer · VMock
Led small teams and built new client-facing features.
Built frontend and backend foundations for Knowah, an online school platform: reusable React components against the design system, and TypeScript introduced for type safety.
Jun 2017 — Jan 2021 (3 yr 7 mo)
Software Engineer · VMock
Frontend engineering, performance work and interface delivery.
Reworked the Webpack build with code splitting, tree shaking, caching and lazy loading. Build times down 20%, initial bundle down 15%.
Built candidate-side modules (Appointments, Events, Interviews) from scratch to a production-ready 8-week deadline.
