What I’m doing now

Location

Singapore, just back from a trip to India.

Work

Started at ST Engineering’s R&D division as a Senior AI Engineer earlier this month. Still in the settling-in phase — meeting people, mapping the landscape, figuring out where I can be most useful. Will write more here once the work crystallizes.

What I’m thinking about

• The agent stack. Which abstractions hold up in production and which are scaffolding pretending to be primitives. MCP feels like it’s becoming load-bearing faster than people expected.
• Fine-tuning as a first-class skill again. RAG-everything was a phase. For domain-specific work, fine-tuning a 7B–13B model often beats prompting a frontier one — and the tooling has finally caught up.
• The research-to-product gap in AI is narrower than it’s ever been, but the engineering around inference, evaluation, and reliability is still where most of the time actually goes.

What I’m learning

• The internals of vLLM and Triton — wanted to understand serving infrastructure beyond “you put a model behind an API.”
• Reading more research papers than I did during the Masters, which is funny — the cadence is just different when there’s no exam at the end.

Side things

• Slowly writing up some of the projects from my Masters — the Real-Time Meeting Summarizer in particular deserves a proper post rather than just a GitHub README.
• Starting to post on LinkedIn more deliberately. Trying to find a voice for short technical notes without crossing into LinkedIn-influencer territory.

Not doing

• Chasing every new framework. The half-life of the AI tooling space is too short to be early on everything.
• Side-project-as-launch. The bar for “thing worth shipping” is higher than it used to be.