Independent software studio
Taraki is a small, deliberately selective studio led by an engineer with 11+ years across both enterprise and startups. We take on a handful of teams at a time and own the whole lifecycle—concept, design, build, ship, and run—turning production-grade web, mobile, and cloud products into real software in users’ hands, fast.
Enterprise delivery inside AXA · 6+ years as a technical lead · the same Australian client since 2019
What we do
We are hands-on builders. Every engagement ends in working software running in production, owned cleanly by your team—and we lean on modern AI tooling to get there faster, without cutting corners on quality.
Fast, resilient web products—from customer-facing SaaS to internal tooling. We pick boring, well-understood technology on purpose, keep the data model honest, and make the common path fast—so the thing is still cheap to change a year after launch, when that matters most.
Apps that keep working when the network doesn't. One codebase across Android, iOS, and web where that's the right call, with local-first storage and background sync—so the app is still useful in a basement, on a job site, or somewhere with one bar of signal.
Production infrastructure on AWS done right—serverless-first, cost-aware, secure, and built to scale without a platform team babysitting it.
Zero-to-one product work: turning a rough idea into a launched, revenue-ready product—scoped tightly and delivered end to end.
Your software rarely lives alone. We connect cleanly to what your business already runs on—payments, CRM, CMS, messaging, and the bespoke internal systems that never came with documentation. The real work is in the failure cases: retries, idempotency, and reconciliation, so a dropped webhook doesn't quietly become a wrong number.
Architecture sized to the actual problem. Static-first where the content allows, serverless where it doesn't—so the common request is a CDN edge hit and not a server under load. Our own reference sites pre-render to hundreds of pages and ship with no origin server that can fall over.
We use modern AI tooling to design and build faster, and we bring AI into your product where it earns its place rather than decorates the roadmap. We also take its failure mode seriously and write the checks that catch it: on our own content platform, anything generated is machine-validated against a verified corpus—every figure, date, and superlative has to trace back to a known source, or the run fails and nothing publishes. A prompt is a request; a validator is a rule.
CI/CD, infrastructure-as-code, and automated deployments with real observability—release pipelines that are fast, repeatable, and dependable, so shipping is routine instead of risky.
Production-grade doesn't have to be expensive to run. We build serverless-first—architecture that scales to zero when idle and up with demand as it arrives—so you get production-grade quality without over-provisioned infrastructure or a runaway cloud bill. Great engineering and a lean operating cost aren't a trade-off; they're the point.
How we work
Speed and quality aren't a trade-off when the team is senior and the scope is honest—and because we build serverless-first, what we ship stays lean to run. Our process is built to remove the things that usually slow projects down, and the costs that quietly pile up after launch.
We start by getting the problem right—users, constraints, and what "done" actually means. If we take it on, we're all in.
You see working software in weeks, not quarters. Automated CI/CD and infrastructure-as-code keep releases fast, repeatable, and low-drama, so progress is always real and visible.
Tested, documented, and architected so it holds up in production. Where correctness matters we enforce it in the build itself—checks that fail the release rather than conventions people are asked to remember—so your team can own it confidently after we hand it over.
Who we work with
We only work on software that matters.
We take on only a few projects at a time—products with real impact, not throwaway work. That focus is the whole point: it's how we deliver work we're proud of, quickly, without spreading a small team thin. If we're not the right fit, we'll tell you plainly.
Tell us about your projectExperience
Taraki is led by Theodoro Pelingan Jr.—a full-stack engineer with 11+ years shipping production software, the last 6+ of them as a technical lead owning architecture and delivery. Based in the Philippines and delivering to Australia on APAC-friendly hours—for the same client since 2019.
The same Australian client, since 2019. One relationship carried across two employment arrangements—the contracting vehicle changed, the client didn't. Which is worth less as a statement about loyalty than as a plain one about follow-through: when he takes something on, he is still on it years later.
Enterprise & regulated
Nine months in 2017 employed by the consultancy VISEO and deployed to AXA, one of the world's largest insurance groups. Microservices, test-driven development, Kubernetes, Jenkins CI/CD, React and Redux front ends, unit and end-to-end testing.
Operating, not just building
Functional and technical specs, estimation and grooming, code review, front end through vertically scalable APIs—plus infrastructure as code, CI/CD, and monitoring and alerting.
Together those two support a claim neither makes alone: he has delivered inside a large regulated organisation, and he has owned systems in production—built and run, not merely written.
A long client relationship shows staying power. On its own it doesn't show range—so the products below are the other half of the answer: several of them, in different domains, designed, built, shipped and run end to end. Read the two together. The first says he'll still be there in three years; the second says he can start from nothing and get it live.
References available on request during our first conversation.
Selected work
Consumer products, content platforms that stay accurate at scale, tools for people working offline, and the systems that keep a real-world business running. Each is our own end-to-end work—architecture, build, deployment, and everything after launch. Go and use them; that's rather the point.
Vehicle log and running-cost tracker — for owners who want to know what their car actually costs
A log for servicing, fuel and expenses, paired with a comparison engine for the decision people make before buying rather than after. Every vehicle exists once, in one dataset, and every table, comparison and recommendation is computed from it at build time—so no figure can drift between pages. Range figures name their standard, EPA, WLTP or NEDC, because quoting those interchangeably is how comparison sites mislead people.
An atlas of Earth's natural wonders — for travellers and the naturally curious
355 wonders across 87 countries: what makes each one extraordinary, and what you actually need to know before you go. The category drifts toward myth and this one refuses it—verified geology and ecology only. That standard is enforced in software rather than merely intended: measurements, dates and superlatives that don't trace back to the site's own verified corpus fail the build.
A fruit encyclopedia — for cooks, growers and curious eaters
354 profiles covering what a fruit is, what's in it, what to do with it, and when it isn't safe to eat. Images are where reference sites get caught—wrong species, wrong cultivar, unripe fruit shot as ripe—so every photo was checked against its own entry and rejected when the two didn't match. Attribution is a build gate rather than a convention: no alt text, credit and licence, no build.
A field-service companion — for volunteers working door to door
Territory and household mapping, visit logging and reporting. Offline isn't an error state here, it's the normal one: every device holds a complete local database, and the app is fully functional with no account, no server and nothing uploaded. The hard part was reconciling two devices edited while apart without losing anyone's afternoon of work. Sync is optional, and the server stores opaque blobs it never reads.
Operational systems for a working workshop — American truck specialists, Queensland
Every vehicle through the door is a one-off build with its own specification, parts, sequence and promised date—hard to run well, and easy to run badly. We build and run the software behind that business: the systems that hold a build together from first conversation to handover, and take the repetitive admin out of the day. It runs daily, unattended, and has become unremarkable—which in production is the highest compliment software gets.
It's also the work we can say least about publicly, which is rather the point. How a client's business runs internally stays the client's business.
Start a project
Tell us a little about your project. If it's a fit, you'll hear back from Theo personally—usually within a couple of business days.
Prefer email? hello@taraki.dev