I'm a freelance full-stack developer currently pursuing an MCA, splitting my time between client work and building my own products from scratch. Most of what I build sits at the intersection of two things I care about: solid full-stack engineering, and applying GenAI to problems that don't usually get attention — like making an LLM+TTS pipeline actually pronounce Burmese correctly, not just technically "support" it.

I think the most honest way to learn is to build something real, hit the actual walls (broken payment gateways for your own country, a TTS engine with no phoneme control, an RSS feed that silently goes stale), and document the fix instead of pretending it went smoothly. That's the philosophy behind everything below.

🌅 AI News Agent — Burmese Morning Briefing An AI-generated audio news pipeline that delivers tech, business, and football news in natural-sounding Burmese every morning. Currently live via Telegram, with a subscription Android app in active development.

• Multi-source resilient news aggregation (17 global outlets, parallel fetch, fault-isolated)
• Self-correcting LLM scriptwriting — an automated pass catches and fixes choppy, unnatural phrasing before it ever reaches speech synthesis
• A hand-built pronunciation-correction system, because getting a neural voice to say "AI" or a footballer's name right in Burmese script turned out to be a real engineering problem, not an afterthought
• Segment-based audio splicing for tonal variation across content types
• (Full case study in progress — the honest version, including what broke and why)

🏙️ Amara Garden City — a real-estate marketing site for UD Group, built with Vite, React, TypeScript, Three.js, GLSL shaders, and GSAP for scroll-driven 3D storytelling.

🗂️ PanThaZin.net — a Next.js/Prisma/PostgreSQL system with role-based auth and a full audit trail.

Frontend

Backend

GenAI / LLM

Tools

• 📚 Studying Network Science & Graph Theory as part of my MCA. coursework
• 🎯 Working toward a Forward Deployed Engineer role over the next 6 months
• ✍️ Creating Burmese-language AI literacy content — because the language barrier to understanding AI shouldn't be as wide as it currently is
• 📫 Reach me at nickytwist70@gmail.com

If you're building for a market or a language that the big platforms treat as an afterthought — the tooling gaps are real, but so is the opportunity. Half of what I've learned building the Burmese briefing pipeline came from hitting a wall that had zero Stack Overflow answers for my exact case (Myanmar payment gateways, Burmese TTS pronunciation, sign-format quirks in a library's validation regex) and just... reading the source code myself. You don't need permission to be the first person to document the fix.