Hi, I'm Donavan.

Welcome to my portfolio. I like tinkering with and building things, especially GenAI applications, software intelligence, and data products that feel useful instead of being merely impressive.

I have worked across applied research, development of GenAI tools, and a variety of ML fields such as geospatial analytics, audience segmentation, and traditional ML tasks (e.g. classification and demand forecasting).

Disclaimer that this portfolio site is almost entirely vibe coded, using Claude (via Cline extension on Visual Studio). See the below section for my thoughts on GenAI.

My thoughts on GenAI

I have mixed feelings about AI.

When gpt 4 first came out, it completely changed the project I was working on. Our initial workflow was with OCR, rules based data wrangling, and hours of manual validation to extract structured data from a wide variety of menu images. When we swapped to (then) Gemini, the zero shot results was amazing, and we understood instantly that we had to pivot our workflow to incorporate LLMs more.

I think GenAI is most exciting when it becomes reliably useful, with grounded retrieval and clear frameworks for understanding performance, but I do not believe that it can replace humans yet, especially for complex problems where human judgement is needed, like compliance, business recommendations, or even just the nuance of understanding what someone does at work.

Fast forward to now, GenAI is used pretty much used daily for coding tasks, including building this portfolio site. At the risk of forgetting fundementals, I still think it is pretty neat that I, as a non software engineer, can build this site and focus on the creative aspect.

However, as someone who still came from a computing background before ChatGPT and LLMs, I wonder if this is the right direction, for people who build systems to not understand even the basics of the systems they are building.

It feels liberating but strange that I can create this site almost without touching the code. (I still did the deployments and read CSS/javascript elements to make my prompting more effective).

Who knows what the future of computer science and coding is lol. This is written in 2026, let's see how well it ages...

Tech Skills

GenAI

LLM applications

Agentic AI

RAG Chatbots

Multimodal processing

Evaluations

Prompt Engineering

Machine Learning & Data

Python

SQL

Web Scraping

Data Processing

Natural Language Processing

Computer Vision

Cloud & Product

Google Cloud

DevOps

Git

Docker

CI/CD

Stakeholder Management

When I'm not coding

I actively take part in triathlons and trial races. I've travelled to Vietnam, Hong Kong, and Malaysia for races.
I make coffee. I believe that it is easier to make a nice drip coffee than an espresso based coffee, despite what others say.
I recently started gaming with my friends. I never gamed a lot but figured I do not have many years left to enjoy gaming with friends. I learnt that I am terrible at games.