Academics · Computing
Why computing
The question I keep coming back to is not whether software can be built, but how anyone knows it works. At OCBC that meant giving automations shared logging and a health dashboard, so a failure shows up instead of quietly producing a wrong report.
Record
Academic record
- Qualification
- Diploma in Information Technology (Software and Applications)
- Institution
- Singapore Polytechnic
- Period
- 2024 – present
- Expected graduation
- April 2027
- Cumulative GPA
- 3.55 / 4.0 (after Year 2, March 2026)
- Certificates
- Certificate in Entrepreneurship (three elective modules)
GPA by semester
3.09
Year 1, Sem 1
3.70
Year 1, Sem 2
3.98
Year 2, Sem 1
3.46
Year 2, Sem 2
Coursework
- Fundamentals of Programming 1 and 2
- Java Programming
- J2EE Application Development
- Front-End Web Development
- Back-End Web Development
- Database Systems
- Secure Coding
- Software Engineering Practice
- Continuous Integration and Continuous Delivery
- Design for User Interaction
- Fundamentals of Computing
Results
Selected modules
Computing modules
- DIST
Database Systems
LocalLoco’s location queries run in PostGIS rather than application code.
- A
Continuous Integration and Continuous Delivery
CI now gates every change to ROUTR and to this site.
- A
Back-End Web Development
- A
Secure Coding
StartupLink enforces access in database rules, not the interface.
- A
Fundamentals of Programming 2
- A
Java Programming
- A
Front-End Web Development
The argument
What I want a degree for
The same question runs through my own projects. ROUTR is measured against a 153-prompt evaluation set with a published method. DoorLens ships with a limitations document because its dataset was too small to claim more. Nudgio is an alarm, so it has to fire on time even after the operating system kills the app.
A diploma taught me to build. I want a degree to learn the parts I keep reaching for and do not yet have: the theory behind reliable systems, the mathematics behind evaluating models, and the discipline of proving rather than testing.
Evidence
The same question, four times
Each one links to the work itself.
Observability at OCBC
Shared logging and an automation-health dashboard that classifies every run as succeeded, failed or stalled.
See the evidenceROUTR: measured routing
A 153-prompt evaluation set, CI validation, and a published method so the accuracy figure can be reproduced.
See the evidenceDoorLens: honest limits
Frozen, hashed models and a written limitations document for a 110-cycle dataset with no test labels.
See the evidenceNudgio: reliability by design
One OS alarm registered at a time, 21 architecture decision records, and a spec with requirements traceability.
See the evidence
Next
What I want to study
Software engineering and reliability
How to design, test and observe systems so that failures are found by the system, not its users.
Evaluating AI systems
Building evaluation sets and methods that say honestly how well a model or agent performs, and where it does not.
Beyond the classroom
Committees and teaching
Leadership roles at Singapore Polytechnic, plus the workshops and hackathons I teach at.
Organising committee
LEAP 2026, School of Computing
2025 – present
Assistant Quartermaster
SPCyclists main committee
2025 – present
ExCo (Events)
Freelance Academy, Media Arts and Design
2024 – present