Curious enough to go deeper. Practical enough to ship.
I'm Twissa, a machine learning engineer, software engineer, and M.S. AI student in Singapore. My work lives where models meet messy data, production constraints, and real people.
A model is only the beginning.
I care about the whole path from an idea to a useful outcome: whether the data is honest, whether the evaluation means anything, whether the system can be operated, and whether the person using it knows what to trust.
That perspective comes from moving between production software engineering and applied machine learning. At Microsoft, I worked on Windows product telemetry and backend services used at significant scale. I'm now a Software Engineer Intern working on AI agents at Bosch, including natural-language data analysis and agent interoperability. Alongside that work, I'm studying AI at NTU and building projects across forecasting, computer vision, and language models.
This portfolio is a working record of that growth. It will get better as the work gets better.
The useful overlap.
The strongest work happens when technical depth and product judgment are allowed in the same room.
Production instincts.
Service boundaries, cost, telemetry, deployment, and the quiet details that keep software dependable.
Applied rigor.
Leakage-aware splits, calibration, interpretability, and evaluation that respects the actual problem.
Clear surfaces.
Writing and interfaces that help people understand a system before they decide to trust it.