Software Engineer  /  Hong Kong

Darren
Wong

I build full-stack web applications and machine-learning systems — from the database schema up to the model evaluation.

01  About

I spend most of my time somewhere between backend systems and machine learning.

I am reading Computer Science at HKUST with an extended major in Artificial Intelligence. Right now I am an AI engineering intern at CarryAI, building a web application that helps people write better prompts for vision-language models — which in practice means a lot of thinking about retry logic, fallback parsing, and what an LLM client should do when the model returns something unexpected.

Before that I spent a summer at Total Rehabilitation Management working on a workflow analytics dashboard and a document classification pipeline, and an exchange semester at UNSW Sydney benchmarking six computer-vision architectures against each other on an agricultural pest dataset.

The problems I like are the ones where the hard part is not writing the code but working out what the system should actually do — schema design, evaluation that is not quietly misleading, the failure modes nobody wrote down.

Outside of that: aviation, finance, and badminton.

02  Selected work

Full source for each
is on GitHub.

01

Library Management System

Catalogue, borrowing and returns, reservations, fines, and role-based access for members, librarians and administrators. Built as a six-person team project, then substantially reworked afterwards: JWT authentication, Flyway-managed schema migrations, and an integration suite that runs against a real PostgreSQL container rather than an in-memory stand-in.

Java 17 Spring Boot Spring Security PostgreSQL Flyway Testcontainers React 18 Vite Docker

169

tests passing

3

user roles

02

Crop Pest Detection Benchmark

Six deep-learning architectures — three detectors and three classifiers — trained and evaluated head-to-head on the Kaggle crop pest dataset, with an end-to-end pipeline covering YOLO-to-COCO conversion, early stopping, explainability visualisations, and a deliberately imbalanced split to test robustness. Completed on exchange at UNSW Sydney.

PyTorch ViT YOLOv11 Faster R-CNN EfficientDet ConvNeXt EfficientNet

93.4%

ViT accuracy

.825

YOLOv11 mAP@0.5

03

StockAgent

Multi-agent LLM stock market simulation · private repository

Autonomous trading agents with portfolio state management — stocks, cash and debt tracking, loan repayment with interest, and bankruptcy detection — trading against an order-matching engine with buy and sell queues and session-based price updates. Four LLM providers sit behind one interface, with format validation, retry handling, and per-agent chat history that carries trading context across simulated days.

Python OpenAI Gemini Claude Ollama

4

LLM providers

04

Safe Reinforcement Learning with Runtime Shielding

A Q-learning agent wrapped in a learned safety shield that vets each proposed action before it reaches the environment, so unsafe moves are filtered at runtime rather than merely penalised after the fact.

Python Keras Q-learning

100%

shield test accuracy

03  Experience

Jun 2026 – Present

AI Engineer Intern

CarryAI Company Limited, Hong Kong

  • Built a full-stack web application that guides users through a multi-turn conversation to generate optimised system and user prompts for vision-language models, with REST endpoints backed by a factory-pattern LLM client abstraction, exponential-backoff retries, and JSON extraction with fallback parsing.
  • Integrated OpenRouter’s Qwen reasoning model with native thinking-token support so the application surfaces the model’s internal reasoning in real time, and built a dual-compression image pipeline — base64 multimodal encoding, MIME detection, presigned URL delivery — that cut payload sizes without visible quality loss.
Python REST APIs OpenRouter VLMs

Jun – Aug 2025

Software Engineer Intern

Total Rehabilitation Management (TRM HK), Hong Kong

  • Engineered a workflow analytics dashboard with data aggregation, multi-criteria filtering and interactive visualisation, letting stakeholders monitor, track and export K2 process metrics across both active and completed cases.
  • Architected a document intelligence pipeline using TF-IDF vectorisation that reached 95% classification accuracy on unstructured document types — incident reports, medical certificates, receipts.
  • Cut regulatory-report image payloads by up to 89% with lossless compression, keeping within third-party platform constraints, and hardened public endpoints with reCAPTCHA verification.
C# EF Core SQL Server TF-IDF ImageSharp

Jun 2025 – Present

COMP & CPEG Mentor

HKUST Department of Computer Science and Engineering

  • Advise first- and second-year computer science students on study planning, academic strategy and co-curricular involvement.

Nov 2024 – Present

Peer Companion

HKUST Counselling and Wellness Centre

  • Facilitate small-group activities supporting social skills and peer relationships, and organise university-wide wellness events.

04  Skills

Languages

Python Java C++ C C# JavaScript TypeScript SQL Bash

Web

Spring Boot React Vite REST APIs HTML CSS

ML & Vision

PyTorch TensorFlow Keras scikit-learn NumPy Pandas NLTK

Data

PostgreSQL MySQL MariaDB SQL Server DBeaver

Tooling

Git Docker AWS Postman Swagger Jupyter

05  Education

Sep 2023 – Jun 2028

Hong Kong University of Science and Technology

BEng Computer Science, Extended Major in Artificial Intelligence

AI/ML · Big Data Mining · Computer Vision · Software Engineering · Operating Systems · Data Structures and Algorithms

2025 Term 3

University of New South Wales, Sydney

Exchange semester, Computer Science

HKUST Fund Exchange Scholarship recipient

06  Contact

I am looking for software engineering internships for Summer 2027, and I am always glad to talk about backend systems, computer vision, or anything aviation-adjacent.

darrenwongsj@gmail.com