About Me
Hi! I’m Khoa Nguyen, a senior researcher and engineer specializing in AI and Privacy-Preserving Technologies.
Background
I finished my Master Degree at Tampere University in Finland in 2021 with a focus on AI and Cybersecurity where I developed a deep interest in building systems that enable machine learning while protecting user security and privacy.
Current & Past Work
I am a senior software engineer at OpenMined, where I build AI evaluation and privacy-preserving machine learning technologies. Currently I’m focused on:
- ScreamingFace: An open toolkit for composing model fusions — several models answering the same question, reduced to one answer — and scoring them honestly against real benchmarks. Every run compiles to a single short line of text that anyone can replay, and model calls are cached and shared, so verifying someone else’s result costs minutes instead of a budget. A result is only worth as much as your ability to rerun it.
Previously I worked on:
- SyftBox: An open-source privacy-first, offline-capable AI infrastructure that enables developers to build applications, run federated learning, and advance Privacy-Enhancing Technologies (PETs) research.
- syft-flwr: A privacy-preserving distributed computing framework that combines Syftbox and Flower. Enable federated learning, analytics, and federated RAG with secure job submission, approval workflows, and offline capability.
- syft-crypto-core: End-to-end encrypted communication and file synchronization for SyftBox using the Signal protocol.
- PySyft: OpenMined’s flagship library for remote data science. It lets a researcher get answers from data they never see or copy, combining differential privacy, federated learning, and secure enclaves.
I have also been researching on privacy-preserving machine learning and AI Safety. You can find my humble scientific contributions on Google Scholar.
AI Safety
I’m interested in AI Safety, especially evaluations, zero-trust system deployments, and AI security. One example is Blindfold, where a model owner, a benchmark owner, and an auditor run one evaluation together inside a secure enclave: nobody has to trust anybody, no secret leaves its owner, and only the score comes out.
Interests
- AI & AI Safety
- Applied AI Systems
- Cryptography & Cybersecurity
- Distributed systems
- Programming Languages
- Boxing: I just had my pro boxing debut fight on 25th October, 2025. Here’s the link to the fight if you are interested in watching
- Guitar & Piano
Connect
Please feel free to connect with me on GitHub, LinkedIn, or X!