The PM who ships it himself, from discovery to deploy.Engineer who happens to own the roadmap.
Engineer-turned-PM with five years in tech. I came up writing production code, so I own products end to end, from discovery to launch, and partner with engineering and design as an equal. Pure PMs can't write the RAG pipeline; pure engineers don't run GTM. I do both.Five years in tech, engineer first. I hardened production systems, led backend delivery, and still ship end to end — RAG pipelines, Flutter apps, n8n automation, CI/CD. The PM title just means I also decide what gets built.
Products I led end to end inside a company: roadmap, metrics, and cross-functional delivery.Platforms I architected and shipped inside organisations — data models, pipelines, and the delivery that keeps them alive.
An insights platform that replaced static PDF reports with interactive, role-based data exploration.Role-based analytics platform that retired static PDF reporting; now growing a natural-language RAG layer on top.
At Varahe Analytics, leadership and field teams ran on static PDF reports: slow to produce, impossible to interrogate, and stale the moment they shipped.
Interactive data exploration with role-based access control, replacing a static reporting pipeline. Prism AI adds a natural-language layer built on a RAG stack — retrieval and grounding over live report data.
Decision-makers could not slice the data themselves. Every new question meant another report request and another wait, so insight stayed bottlenecked behind a document.
Owned the architecture and delivery end to end — data model, role-based views, and the release plan. Aligned engineering, data, and field teams on one roadmap and got Phase 1 into production in six months.
I led PRISM end to end, building interactive, role-based data exploration. I aligned engineering, data, and field teams around a single roadmap and shipped Phase 1 in six months. I now lead Prism AI, a natural-language layer built on a RAG stack.
Static PDF reports fully retired; decision-makers query the data directly. Prism AI, the RAG-backed natural-language layer, is now in flight.
A constituency-scale data platform built from scratch under a hard, immovable deadline.Constituency-scale voter data platform plus a custom LMS, built from scratch against an immovable deadline.
For the South-West Graduates Constituency MLC election, a first-time candidate needed a data and outreach platform serving 85,000+ voters. Nothing existed.
A custom LMS for the calling team, plus an automated pipeline linking training data to measured caller output and call quality — real-time dashboards over fast-moving, messy field data.
Fast-moving, messy data had to become real-time insight, and a calling team had to be trained and measured, all before a fixed election date that would not move.
As Technical Head I owned everything from data design to field rollout, shipping the whole system from zero under a fixed election date that would not move.
As Technical Head, I owned end-to-end delivery across engineering, field ops, and leadership. I built a custom LMS for the calling team and an automated pipeline that linked training to caller output and quality, so impact showed up directly in the data.
85,000+ voters served, real-time insight in the hands of the campaign, and a first-time candidate over the line.
Founder-level builds I designed and shipped myself, from first commit to launch.Products where I wrote the first commit and the last deploy — solo, full-stack.
A context-aware RAG assistant with source-grounded answers, built end to end.RAG assistant with source-grounded answers — semantic chunking, filtered retrieval, and citations — built solo.
A founder build: an AI assistant that needed to answer reliably across a sprawling, heterogeneous body of source material.
Flutter client on a Firebase backend. Ingestion pipeline over 100+ heterogeneous sources, semantic chunking with structure preserved, and retrieval tuned before the model ever sees a token.
Generic LLM answers were not trustworthy; responses had to be grounded in real sources and stay accurate as the underlying material kept growing.
Solo, end to end: data pipeline, retrieval layer, client app, and deploy. Answer quality moved on chunking and retrieval, not prompt engineering.
I built a context-aware RAG assistant that returns source-grounded answers, backed by a data pipeline ingesting inputs from 100+ sources, shipped on a Flutter + Firebase stack.
A live assistant where every answer cites the source it came from, staying accurate as the underlying corpus keeps growing.
A coach dashboard platform built hands-on with the founder, so she could scale past manual ops.Full-stack coach dashboard — clients, scheduling, programming — shipped fast and iterated against a real user's workflow.
A solo coach was running her whole practice by hand. Scheduling, client tracking, and programming lived across spreadsheets and DMs, which quietly capped how many clients she could take on.
A full-stack platform: client management, scheduling, and programme delivery in one app, replacing an ops layer that lived in spreadsheets and DMs.
Growth was bottlenecked by operations, not demand. Without one place to manage clients and delivery, every new client added overhead instead of leverage.
Designed and shipped end to end, working 1:1 with the founder — a tight loop of shipping against her live workflow and iterating on what real usage surfaced.
I built the coach dashboard end to end, working directly alongside the founder: a full-stack platform to manage clients, schedules, and programming in one place, shipped fast and iterated against her real workflow.
Manual spreadsheet ops replaced outright, and a practice that scaled past its previous client cap.
Self-hosted thinking on building products, shipping under pressure, and the move from code to roadmap.Self-hosted write-ups from the trenches: retrieval, automation, and shipping under hard deadlines.
I started as a software engineer, hardening production systems and integrating APIs, then led backend delivery as a Team Lead, setting the coding standards, QA, and CI/CD that a product runs on.
That foundation is why I move fast as a PM. I own my product area independently, define the metrics that tell me whether it worked, and partner with engineering as an equal. Alongside that I take on founder-level builds, from a coach dashboard to a RAG assistant, and advise startups on product, GTM, and automation.
Software engineer first: hardening production systems, integrating third-party APIs, then leading backend delivery as a Team Lead — owning the coding standards, QA gates, and CI/CD pipelines the product ran on.
That muscle never went away. I still take products from first commit to deploy: a RAG assistant on Flutter and Firebase, a full-stack coach dashboard, automation pipelines in n8n. The PM title just means I also decide what's worth building — and I can read every pull request that builds it.