Jam Tic Bong
AI Pipeline
Data Collection to Deployment
Frontend
Technology for Better Lives
Design
Personal Touch & Storytelling
AI Pipeline
Frontend
Design
Production ML and a Bit of Design Curiosity
Production-focused ML Engineer building end-to-end deep learning systems for noisy, real-world data — from forecasting and computer vision to FastAPI deployment.
设计 · 还有一点好奇心
End-to-End
Solutions
I ship production ML and full-stack products for noisy, real-world data
Online
Hybrid
On-site
Freelance
Full-time
Full-time
Featured work showcasing production ML systems, computer vision pipelines, and full-stack products shipped for real clients and industrial partners.
Wabi Sabi by Ouktey — Nail Studio Booking
A bilingual booking site for a one-chair nail studio in Phnom Penh. The public page and the studio desk read the same bookings. Ouktey needed a site that felt like her studio, not a booking widget. I designed and built the public page and a private desk. Guests browse sets, pick a shape and length, choose a day and an hour, and review the request before sending it. No account needed. The studio calls to confirm, and changes go through Telegram. The desk runs the day. Bookings, prices, the promo, the menu, and closed hours live in one record, and anything saved there shows up on the open site straight away. Google Calendar stays in step with the book, so cancelling an event frees the hour. The page is built for phones in Cambodia first. What it does: • English and Khmer copy, with prices and phone numbers kept in Latin digits • Shape and length picker that carries through into the booking • Four slots a day, 10am–6pm, with a two-hour lead time; passed hours stay marked • Review step before anything is sent • Service menu, client clips, studio map, and a promo that can be switched on from the desk • Admin desk for bookings, content, and schedule changes • Two-way Google Calendar sync, plus a Telegram message to the studio when a booking is created, moved, or cancelled • Scheduled ping so the free-tier database doesn't pause after a quiet week
FemFit — Real-time E-commerce Platform
Mobile-first storefront with catalog, cart, checkout, and admin panel (Next.js 15, React 19, TypeScript, Drizzle ORM, Supabase). Modular monolith with cart, orders, payments, and admin modules; Row-Level Security, ABA PayWay with HMAC-SHA512 webhook verification, and transactional stock reservation for concurrent checkouts.
RECS Production Forecasting Systems
Production deep-learning forecasting across three domains at RECS Innovation: (i) household energy-consumption prediction for Self-DR-VPP using KESIS data (LSTM / seq2seq); (ii) KOSPI 200 Index price forecasting with attention models and walk-forward backtesting; (iii) construction bidding-price estimation via stacked quantile regression / RF / XGBoost with wavelet denoising. Shipped as FastAPI services with CI/CD and documentation.
Santi Domnak Rentals
Marketing and enquiry website for a Siem Reap rental business serving students, workers, and expats — designed and shipped as an independent contractor.
View All
Skills(4)
Skills(4)
Skills(4)
Skills(4)
Production ML Systems
End-to-end deep learning for noisy real-world data — time-series forecasting, hybrid ensembles, evaluation protocols, and FastAPI deployment.
Computer Vision
Vision pipelines from data capture and annotation through training, evaluation, and on-device inference for industrial and municipal use cases.
Full-Stack Engineering
Production web apps with Next.js, Vue, TypeScript, Supabase/Postgres, payments, and secure order pipelines — from marketing sites to e-commerce.
MLOps & Delivery
Shipping models as maintainable services with CI/CD, documentation, backtesting, and handoff for non-ML engineering teams.
FAQ
FAQ
FAQ
FAQ
What do you specialise in?
Production ML engineering — time-series forecasting, computer vision, and shipping models as FastAPI services — plus full-stack delivery with Next.js, Vue, and TypeScript for real products.
What kinds of AI projects have you shipped?
At RECS I delivered household energy forecasting (Self-DR-VPP), KOSPI 200 price models, and construction bidding-price ensembles. At Netvision I built the vision pipeline for a smart waste-collection system that won LINC 3.0 First Prize.
Do you work remotely and internationally?
Yes. I collaborate across time zones and have shipped work in South Korea, Cambodia, and remotely. I am currently based in New Zealand and open to remote or hybrid engagements.
What is your development process?
Agile delivery with clear evaluation metrics: discovery and data understanding, modelling and backtesting, production services and CI/CD, then documentation so non-ML engineers can maintain the system.
Let's CreateSomething Exceptional
Open to production ML roles, research collaborations, and full-stack engagements — currently based in New Zealand.