Quan Ton

AegisMCP is a CLI-first, air-gapped-by-default multi-agent workspace for private financial documents. PDFs stay on your machine. Agents talk to tools only through an allowlisted gateway. Outputs are typed RiskReport JSON + Markdown — with evidence citations — not free-form chat spam.

Personal portfolio website for Bennedict Quan Ton — Aspiring Agentic AI Engineer and undergraduate @ UTS & HCMUT. The site showcases projects, education, career journey, and personal interests, with a focus on agentic workflows, multi-agent systems, RAG pipelines, and high-performance AI engineering.

Multi-agent AI counselor for HCMUT admissions, by chat and voice. Every answer is grounded in official data: figures come from typed SQL tables, policy from hybrid retrieval, and a deterministic verifier checks each number before it reaches the student.

A public-benefits-style application form that a screen-reader user, motor-impaired user, or non-native English speaker can complete entirely by talking to their AI agent — because the page itself declares real, callable actions via WebMCP, instead of the agent guessing at a UI it can't reliably see.

Track 3 - LLM Inference Optimization Challenge

The Lantern is a multilingual AI maître d' that takes a full table's order by voice, remembers every word, never invents a dish, and keeps the kitchen in perfect sync. Powered by AssemblyAI, running on local AI.