I build software you can use without even thinking. I have worked at large companies like Traveloka and at startups, so I know how to right-size software to what you actually need, and I train your team to run whatever I build.
What I offer: AI assistants and agents, connectors that put your existing tools inside Claude or ChatGPT, websites and web apps, custom software and MVPs, and advice on AI strategy and training.
Right-sized builds and advice, shaped with you along the way.
I iterate closely with you, so the result fits your needs.
Your feedback and my judgment finish it.
I start from the result your business needs, work out what your users are really trying to do, and build on strong software engineering fundamentals. You get software that moves the numbers you care about, that people actually use, and that holds up as you grow.
I have worked at large companies like Traveloka and at startups, so I know what heavy process costs and what a lean build can skip. I currently work at a GenAI startup in Southeast Asia, which keeps me close to how AI is actually built and used.
With a bachelor's and a master's in computer science and years of building real systems, I am comfortable in any language or stack. Syntax was always the easy part. The fundamentals are what last, and they matter even more now that AI can write the code. So I choose the tools that fit your problem, not the ones I happen to know.
AI does not replace good UX or the way people actually work. In the right hands, it lets me build great software with far more range than before. We are in a golden age of software, and the skill that matters most is knowing what to build and how people will use it.
AI changes every few weeks. Following it is part of my job, so it does not have to be part of yours. I try new tools and models, keep what proves useful, and bring you only what fits your business.
Tech-agnostic by design: I choose what fits the problem, and I have built with all of these.
Product names are trademarks of their owners. Listing them does not imply affiliation, partnership or endorsement.
A schedule we agree at the start, so you always know where things stand.
I handle how it is built. You decide how it looks, how it helps people and what it costs.
Documentation, hands-on training and a clear handoff, so your team can run it without me.
Work done for employers and clients. Names and private details are kept out.
Around every holiday, two or three people in HR chased 30+ client companies and 100+ employees for overtime sign-off, one email at a time. Replies got lost and approvals came late, which delayed payroll and billing.
I showed a working demo within two weeks, then iterated with HR and tested for a week, so it was live in six weeks. I didn't buy an extensive, expensive suite. I built a small system that fits the tools HR already used, with secure expiring links and no sensitive data in emails.
Around every holiday, HR used to chase 30+ client companies and 100+ employees for overtime sign-off. Now it is fully automatic: set up once, it sends the requests, collects the answers and records the approvals on its own. It recovers from problems and delivers approvals in minutes, and a human can still step in and correct it.
Recruiting ran on a Google Form and one job board. Applications went to a spreadsheet and were sifted by hand by two or three recruiters. Candidates waited when someone got busy, and in a BPO, every day a candidate waits is money the business loses. About three or four were screened a week.
I fixed only what was broken, the sifting, and kept the Google Form and job board. I set the foundation first (architecture, security rules, infrastructure, CI), then split the work into three teams with clear contracts between them. I steered the teams toward cheap and fast choices. The developers' AI agents worked from written specs, and we held regular live demos to ensure quality.
One searchable place holds every applicant. Resumes are read automatically, and recruiters work through stages from applied to hired with shared notes. Analytics show where hiring stalls. Recruiters now screen about three times as many candidates a week.
The client had a promising startup idea for youth sports, but could not show potential investors and users how it would work. Thousands of young athletes and hundreds of programmes had no shared place to connect: a recruiter cannot visit every game, a coach cannot reach at scale, and an athlete cannot get noticed.
I built it from scratch to deployed in a month and a half, with design input from a designer. I scoped it deliberately: exports and a full chat system stayed out, which kept the timeline realistic. Environments, CI and monitoring were in place from the start, and demo accounts and flows were ready for the pitch.
A working platform prototype for six kinds of users, from athletes and parents to coaches, organization admins and college recruiters. It has public profiles, verifiable combine results, video review with annotations, self-managing subscription billing, parental controls and an admin panel. The owner could show it to investors, partners and customers, and testers and some real users tried it.
Sample builds made to show what I can do for your business.
A demo clinic site where booking, payment and an assistant grounded in the clinic's own documents share one flow.
My own projects, with live links where they exist.
See more on GitHub
NAPAS JakartaA live Jakarta air-quality map with an assistant that answers questions from monitor readings, free for everyone as a public good.
ServersUpA live status board for World of Warcraft and Final Fantasy XIV realms, plus a Discord bot that reports server status and alerts your channel the moment a server goes down.
I design and build enterprise AI applications at an early-stage startup in Southeast Asia, turning fast-moving GenAI tools into software that businesses can rely on. I work closely with users to build what actually helps.
I helped keep the systems that bring in inventory from more than 70 partners running reliably, at over 50 million updates a day. I redesigned how the platform connects to partners, so new ones could be added in about half the time, and delivered three major partner integrations on schedule. I also mentored three junior engineers.
I developed a cost-effective, real-time alert system across hundreds of IoT devices for a school-bus optimization SaaS startup. When incidents like breakdowns, crashes or route changes happen, parents and school administrators are notified within 3 seconds, and a detailed report follows in under two minutes. This was crucial to establishing a working relationship with a critically important enterprise client.
Between 2019 and 2022 I completed four software internships and research roles at Traveloka, UMass Amherst, MITRE and Constant Contact. I built a full-stack research application for repeated 200+ person studies in semantic query construction, turned COVID-19 and socioeconomic data into insights on vaccine uptake alongside a public-health expert, and built tools that helped engineering teams test and process data faster.
University of Massachusetts Amherst. Coursework included advanced data structures and algorithms, distributed databases, and distributed operating systems and systems.
University of Massachusetts Amherst. As an undergraduate, I was credited as a contributor to SQuID, a UMass Amherst database research project on discovering query intent from examples.
Statement of accomplishment from DataTalks.Club for successfully completing the LLM Zoomcamp.
Use Get in touch at the top to pick a free 30-minute slot, or skip the call and email me. We will talk through what you are trying to do and whether I am the right person to help. No obligation.
Email me at inquiries@armasn.dev