Banking & Payments Technology Leader · AI Strategy, Agentic Systems & Governance

For 20+ years I've built technology inside high-stakes financial systems — credit cards, commercial payments, tokenization, clearing, disputes. Systems where a mistake moves real money, so every action needs identity, authorization, evidence, and audit.

Now I'm applying that discipline to a new question:

How do we put AI in front of students and classrooms without giving up privacy, judgement, or trust?

Payments taught me that trust is engineered, not assumed. AI is no different.

An open source AI marking assistant for Singapore exam essays — PSLE composition and A-Level General Paper, with more exams coming. Marks against each paper's band rubric, underlines every error with corrections, coaches improvement from the student's own writing — and it's private by default: runs fully local, so no essay ever leaves the machine.

Teachers get their evenings back. Students get unlimited, patient, specific practice. Nobody's child becomes training data.

Free & open source (MIT) · local models to frontier models, your choice · ships with its own eval set

• A prompt is not permission. Intent isn't authority — production AI needs identity, policy, approval, audit.
• Evals over demos. Every repo ships with its own evaluation set. If you can't measure it, you're marketing it.
• Local-first where data is sensitive. Children's writing is the clearest case there is.
• Augment the teacher, never replace the judgement. The AI does the tireless part; the human does the human part.

Built one repo at a time. Each release ships working code, honest evals, and a written argument. On the bench: verification-first RAG · exam generation from any document · a public benchmark for AI essay marking.

✍️ baijayanta.medium.com — AI strategy, agentic systems, payment infrastructure 💼 linkedin.com/in/baijayantaroy — where I publish most

📍 Singapore · Views my own, not my employer's.