Agentic AI · scalable systems · built to connect
I build agentic AI, make it adoptable, and engineer the systems it runs on.
For teams shipping agentic AI that has to survive production, and founders building a product that has to scale. The two capabilities connect: agents need the platform to run on, and the platform needs the intelligence to be worth running. Proof in capital markets, legal AI, and scale-ups.
Currently taking a small number of consulting engagements, alongside my role.
Proof, up front
“Code that works isn’t enough. It has to work for the business.”
So the work shows up as agentic AI running live in capital markets, legal-AI research that classified 1,200+ judgments for ~$30, and a storefront whose speed stopped costing customers.
Selected work
What I’ve shipped
Recreated mockup — confidential engagement.
Agentic AI for capital-markets infrastructure
I built the autonomous AI agents that triage the exchange and vendor notices a market-data team drowns in. Then the enterprise SSO, idempotent integrations, and audit trail that let a bank's IT and procurement teams approve the rollout. The model is the easy part; making it deployable is.
A self-serve storefront scaled to enterprise tier
Service Express shipped as a new product. Four WorkWave suite integrations. Fan-out notification architecture. TypeScript migration and code-level security work. Plus the ~30x performance gain that stopped the flagship page costing prospects.
Recreated mockup — confidential engagement.
One inbox, every channel
Led a small engineering team owning communications at a $35M-funded proptech with $1B+ under management. Multi-tenant comms across email, SMS, push, and in-app; a Kustomer CRM integration that gave ops one view; a legacy codebase moved to TypeScript without a rewrite.
AI that reads case law
A RAG and prompt-engineering pipeline for Sri Lankan court judgments: 1,200+ documents, an 89-category taxonomy, built and evaluated for ~$30. Published, with an honest retrospective on its own evaluation limits.
Recreated mockup — confidential engagement.
A sales team in your pocket
Field-sales operations rebuilt for agents who lose signal mid-route: offline-first cross-platform mobile (one codebase, iOS + Android), GPS routing, and a PHP-to-Node API-first backend. Intern to architect in four years, and it shipped across Asia.
Engagements
How I help
Three problems I’m usually brought in to solve. Often it’s one. Sometimes the agentic workflow and the platform it runs on need to be built together.
Agentic AI & enablement · lead
AI agents enterprise IT teams will actually approve and deploy
Who: teams shipping agentic products or automation in regulated or high-stakes domains.
Outcome: autonomous workflows that hold up in production: enterprise SSO, idempotent integrations, auditable output, multi-tenant from the first commit.
Starts with → an AI readiness review.
See it in capital-markets agentic AI →Scalable web & mobile
A product built for the next 10x: web and mobile
Who: founders and teams building or scaling a web app or a cross-platform mobile product.
Outcome: a scalable web app or one cross-platform mobile codebase (iOS + Android) on fundamentals that hold under load: multi-tenant from the first commit, fan-out architecture, and auth that enterprise IT teams will approve.
Starts with → an architecture review.
See it in the WorkWave case study →Engineering enablement
Get your team shipping reliably, without the constant handholding
Who: founders whose engineers need levelling up and a delivery process that holds.
Outcome: juniors mentored and the process that makes a team ship reliably: code review, CI/CD, and the delivery discipline I built running a small engineering team.
Starts with → an engineering review.
See it in leading comms at a $35M-funded proptech →What you get
You talk to the person doing the work
The builder and the decider
You talk to the person writing the code and making the architecture calls. Nothing gets lost in a handoff, because there isn’t one.
AI that’s actually adoptable
The agent is the easy half. The harder half is SSO, idempotent integrations, and audit trails that enterprise IT teams will approve, built in from the start so the work gets used, not piloted and shelved.
Systems that hold as you grow
Built to scale, not just to demo. The capital-markets deployment went through a bank’s infrastructure review because the foundations were right. That’s the kind of architecture that holds at the next 10x instead of forcing a rewrite.
Builder, team lead, published researcher, and someone who’s shipped production agentic AI where the rules are strict. Each is common on its own; having all of it on your problem is the rare part.
What they say
People who’ve worked with me
“He articulates complex architectural decisions in business terms without oversimplification — an engineer who understands both the technical and strategic dimensions of product development.”
“As our Tech Lead, his leadership and technical skills raised our team's standards. His communication helped stakeholders and the tech team connect.”
“In a short time he was working fully autonomously and took over almost all responsibilities for a complex project — a great leader on it. A skilled lead software engineer.”
“Innately curious, detail-driven, and meticulous — he has certainly left a legacy in the time that he helped build.”
“His proactive approach to understanding the business context — he adds significant value with insightful suggestions and optimizations.”
“He constantly asked insightful questions, seeking to understand why we did things a certain way — which often led to improved processes and solutions.”
How I think about the problem
How I approach the work
Why before how
I get the business problem before I write code. The framework is the last decision, not the first.
Humans first
Product, UX, and the stakeholders are in the room early. No surprises at launch.
UX is my problem too
If the flow's bad, I'll say so. Engineers who ignore UX ship garbage.
Shipped ≠ done
I care whether it got adopted, held up under load, and actually solved the problem. Not just whether it merged.
About
I’d rather build the thing than report on it.
I left finance the moment I’d rather build the system than report on it, and I moved fast because the work landed: intern to architect on a product suite that shipped across Asia, then leading a small engineering team at a $35M-funded startup, taking agentic AI into capital markets, and publishing legal-AI research. The constant is that I start with the business problem, build the real thing, and stay honest about how it went. My own research shipped with a frank retrospective on its limits.
Off the clock it’s curiosity over comfort: I play games to understand how good systems are designed, and I’m drawn to anything where the craft shows in the details.
The longer version →800×800 · supporting
Start here
Have an AI, automation, or platform problem that has to work in production?
Tell me what you’re trying to fix, or book a 25-minute call. If I’m not the right person, I’ll say so.
Or book directly
Start with a 25-minute call
Bring the problem, the constraints, and what “fixed” looks like. If I’m not the right fit, I’ll point you to who is.
Discovery call · 25 min →Prefer async? LinkedIn · hello@muljayan.com