Working AI agents, n8n pipelines, and system integrations that live in Slack, WhatsApp, email, and your CRM. Built against your actual workflow, then handed over for you to own.
The shape of an engagement
We look at your actual workflow and find the one bottleneck worth automating first. No audit theatre, just a real decision about what pays off.
I build the system against an agreed scope. Usually n8n, APIs, and a well-placed agent. You see it working as it comes together.
It goes live in your tools, your team starts getting time back, and I hand over something you actually own, not a subscription to me.
Finished systems are the point.
Sometimes the right first step is a 90-minute map of where automation pays off. But the destination is always something working, not a 40-slide report on what could be automated someday.
No custom software unless it's the right answer.
Most of the time, wiring up the right tools does the job. If you genuinely need more, I'll tell you.
I work solo and stay close.
The person who scopes it builds it. You're not handed off to a junior after the first call.
Capabilities
These aren't a service menu, they're the shapes the work usually takes. Most engagements end up a mix.
Slack, WhatsApp, email. Custom AI assistants that answer questions from your internal docs, summarise the day, and stop the constant “where's the file?” messages. Your team leads stop being a help desk.
A lead comes in, lands in the CRM, triggers a follow-up, and logs itself. Reports that took two days build themselves. Built on n8n and clean APIs so data in silos finally talks to the rest of your stack.
FAQs, order status, routing, on the channels your customers actually use, 24/7. Your people handle the conversations that need a human, not the ones that don't.
My own products
Both of these are mine — designed, built, and shipped end to end. They're live right now, not behind an NDA. Go open one. Then imagine what I'd build for yours.
An AI teammate that works while you sleep — and asks permission before it touches anything.
Stop betting your decisions on one AI's opinion.
Proof
Anonymised — the work is real, the names aren't. Each one shipped and used.
What it was
A team was spending hours every week assembling the same status deck.
What I built
A system that pulls the underlying data, assembles the slides, and leaves only the human judgment to humans.
What it did
The weekly deck that used to eat an afternoon now builds itself.
What it was
A marketing team was manually reviewing hundreds of creator profiles.
What I built
An agent that reads each profile, scores it against the team's criteria, and produces a shortlist with reasoning.
What it did
The team reviews the shortlist, not the pile — and the pile was the bottleneck.
What it was
The same questions landing in Slack every single day.
What I built
An assistant that answers from internal documentation, with sources, directly in the channels people already use.
What it did
Repetitive questions get answered instantly. The team gets their focus back.
Something on your workflow look familiar?
Let's talk about it →Everything else in the workshop — utilities, frameworks, and tools, all public. Seepient and OpenFusion above are where this work goes all the way to product.
Field Notes
Build logs, strategy, and industry breakdowns. Specific over conceptual, written to be useful.
The model with the loudest hype is rarely the right hire for every task in your business. Here is how leadership can actually assess AI models: by reliability at your real work, and by honest cost math that could go either way.
Sep 23, 2026A candid comparison between Smol Analytics and Google Analytics 4 on performance, implementation costs, privacy, and how AI agents are rendering legacy dashboards obsolete.
Aug 29, 2026A complete build log of how I implemented Smol Analytics (created by Arjun Patel) on a private VPS, instrumented Zyntopia's Next.js application, and connected AI agents for autonomous monitoring.
Aug 28, 2026A lighter path
I also do short strategy engagements — a working session to map where automation will actually pay off in your business, and just as importantly, what to leave alone. It's the same 15-minute call as a build conversation; if it turns into a scope, it turns into a scope. If it doesn't, you leave with a map.
Book a callAbout
I'm Hashan. I've spent the last few years building AI agents and automation systems in production — the kind that actually get used every day, not demos that look good in a pitch.
I started Zyntopia as a solo practice because I kept seeing the same gap: teams losing real hours every week to work a system could handle. I'd rather ship one automation that does the job than talk about transformation.
I build in the open. Most of what I make lives on my GitHub. If you want to know whether I can actually do this, that's the honest place to look.
FAQ
Operations-heavy teams of roughly 10–50 people: agencies, professional services, growing startups. The common thread is a team losing real hours every week to repetitive work that a system could handle. If that's not you, I'll say so on the call.
With an agency, you pitch a senior partner and get handed off to junior contractors learning on your dime. With me, the person who scopes the architecture is the person who writes the code and puts it into production. You get faster turnaround, no communication telephone games, and direct accountability.
Mostly n8n for deterministic workflow pipelines, open source LLM agent frameworks for reasoning tasks, and direct REST APIs for your stack (CRMs, Slack, databases). I prefer open, self-hostable tools over proprietary black boxes so you retain full ownership of your data and logic.
A single automation (“the fix”) is usually one to two weeks. A connected system across a few tools is more like three to four. We agree the scope first, so you know before we start.
Every engagement is scoped to the actual problem, so I don't publish a price list. As a rough anchor: single automations typically land in the $X–$Y range, and connected systems more. Book a call and I'll tell you honestly whether it's worth doing and what it would run. TODO(hashan): replace $X–$Y with your real band.
No. Everything is fixed-scope, fixed-price. We agree what's being built and what it costs, then I build it. If you want more after that, it's a new project.
Yes, NDAs are standard. Wherever it's avoidable, I build on your accounts and your infrastructure so your data doesn't pass through mine. We'll talk through exactly what touches what before anything gets built.
I don't build custom software when off-the-shelf tools will do the job, and I don't sell open-ended “AI transformation” retainers. If what you need is a strategy report rather than a working system, I'll say so.
Yes. When it's done, it's yours, running in your tools, on your accounts. I hand it over properly, not into a dependency on me.
You own it, so you're not locked into me for fixes. That said, I'm available for follow-up work as a new engagement if something needs changing or extending, and I'll hand over documentation so you or another dev can maintain it without me.
Book a call
Tell me. Book a call and we'll look at your actual workflow together. If there's something worth automating, we'll see it in fifteen minutes.
30 minutes. No prep, no deck, no obligation.
Quick note on what you want automated, then pick a time that works. I read every one of these before the call.
Prefer email? hashanwickramasinghe@gmail.com