I help teams take AI and cloud-native ideas from "this could work" to something running in production: computer vision on edge devices, LLM-backed automation, and the Kubernetes and AWS infrastructure that keeps it all up.
Nisal Palliyaguru10+ years buildingColombo, Sri Lanka, working remotely worldwide
What I do
Most engagements are some mix of these. If you're not sure which one your problem is, that's usually what the first call is for.
Computer vision and LLM features taken from prototype to something your users actually touch.
Backends designed to survive traffic spikes, team growth, and the second year of maintenance.
Deployment that stops being an event, so your team ships on a Tuesday afternoon without flinching.
An honest second opinion before you commit budget to an architecture or an AI roadmap.
How I work
No discovery phase that produces only slides. Every stage ends with something you can read, run, or ship.
A call, then a short written brief: what the system needs to do, what it explicitly doesn't, and where the risk actually sits. If the honest answer is that you don't need me, I'll say so here.
A concrete technical plan covering services, data model, infrastructure, and cost, before any real code is written. You get a document you could hand to another engineer.
Working software in short increments, deployed to a real environment early. You see progress in a browser, not in a status report.
Documentation, runbooks, and a walkthrough with your team. The goal is that you can maintain and extend it without me on retainer.
Selected work
Platforms serving real traffic: national news sites, government infrastructure, election-night dashboards, and marketplaces. Built end to end, from architecture through deployment.
Jan 2024 – Present
Developed a tri-language news platform for Ada Derana 24x7 with scalable web apps, dashboards, and backend architecture. Designed AWS-based infrastructure to deploy test environments, ensuring reliable and seamless content management.
May 2024 – Jan 2025
Designed and implemented a microservices-based platform for civil societies to publish content, appeal for volunteers, and foster collaboration. Led architecture development, implemented GitOps workflows with Kubernetes and Argo CD.
Jul 2024 – Dec 2024
Built an interactive real-time dashboard to visualize Sri Lanka's election data with seamless data updates. Designed an intuitive admin dashboard for real-time reader engagement and efficient data handling.
Dec 2022 – Jan 2024
Developed three news platforms with dashboards for article management, integrating cloud-based backends for real-time updates. Enabled analytics for insights and optimizations. Recognized in BestWeb.lk Awards 2024 with Merit awards.
Mar 2023 – Sep 2023
Developed a video-based classified advertisement platform with an admin dashboard for ad management. Built a scalable backend for reliable data handling, analytics, and performance optimization.
I'm a software engineer with 10+ years building production systems: cloud-native backends in Go and NestJS, real-time platforms on AWS and Kubernetes, and more recently machine learning that runs on real hardware rather than in a notebook. I started freelancing in 2015 and never stopped, which is where most of the 40+ clients came from.
Most of my work has been the unglamorous half of the job: multi-tenant microservices for a government ministry, GitOps pipelines that let small teams deploy without fear, election dashboards that had to stay up on the one night that mattered, and detection models compiled to ONNX so they run on edge devices instead of a GPU cluster.
Technical explorations of algorithms, ML models, and systems powering the products we use every day, with interactive visualizations and real source code.
Coming Soon
How YouTube decides which videos to surface: candidate generation, ranking, and post-processing filters.
Whether you were joining a neighborhood group, a community chat, or connecting with someone you just met at an event, sharing your digits… Continue re...
You don’t need to be a tech person to read this. This is for the school leaver wondering what career to pick. The office worker who keeps… Continue re...
A high severity Server Side Request Forgery (SSRF) vulnerability has been disclosed in Next.js Continue reading on Medium »...
FAQ
Most engagements are either a fixed-scope project (a defined system built and handed over) or a part-time retainer of a set number of days per month for ongoing architecture and delivery work. Short advisory engagements, such as an architecture review or a feasibility assessment, are usually a week or two.
I'm in Colombo, Sri Lanka (GMT+5:30) and work remotely with clients globally. That overlaps comfortably with a full European working day, most of the Middle East and Asia, and early mornings in the US. I've been fully remote or hybrid for most of my career.
Fixed-scope projects are quoted as a fixed price once we've agreed the brief, so you're not exposed to my estimating errors. Retainers and advisory work are billed on a day rate. I'll give you a number after the first call, not before, because a quote without scope is guesswork.
Not very. A one-week architecture review or a single AI feature bolted onto an existing product is a perfectly good engagement. What I'm poor value for is short-notice bug-fixing on a codebase nobody has documented.
No, and it's usually better if you don't. A lot of what I do early on is working out which parts of a problem genuinely benefit from a model and which are better served by ordinary software. That distinction saves more money than any tooling choice.
You describe the problem, I ask questions, and by the end you get a straight answer on whether it's feasible, roughly what it takes, and whether I'm the right person for it. Thirty minutes, free, no follow-up sequence.
Bring the problem, however vaguely defined. We'll spend half an hour working out whether it's worth building and what it would actually take.
30 minutes, free, no pitch. Just scoping the problem.
If a calendar invite feels premature, email or a message works just as well.
