Welcome to my personal space!

This is where I share my research updates, technical posts, travel memories, and stories of the wonderful people I've met along the way. Beyond AI, I am curious about psychology. My work models people as rational, strategic agents, and psychology reminds me how much richer natural intelligence and the human inner self really are.

  Short Biography

I am Linh. I am currently a final-year PhD candidate in the fields of Decentralized AI and Algorithmic Game Theory at Trinity College Dublin, The University of Dublin. My PhD is fully funded by the School of Computer Science and Statistics, Trinity College Dublin and Research Ireland ADVANCE CRT.

My work focuses on Federated Learning/Unlearning paired with Incentive Mechanisms. I design game-theoretic incentive frameworks for organizations, such as hospitals and banks, that train AI models together while sharing only model parameters. These organizations both cooperate and compete, and each keeps its local resources (e.g., data quality, computing power) private from other organizations (information asymmetry). My frameworks help them decide how much to contribute, reward them in proportion to their marginal contributions, and preserve their verifiable right to be forgotten.

  Research Experience

• Jul–Sep 2026Research Intern, Montimage, Paris, France
  Client-level federated unlearning
• Jun–Sep 2025Research Intern, Delft University of Technology (TU Delft), Netherlands
  Incentives for federated unlearning

During my bachelor’s, I led research on Software-Defined Networking, Wireless Sensor Networks, deep learning-based IoT systems, and blockchain-based data sharing. I also collaborated on projects funded by NAVER Corporation, the APNIC Foundation, CSIRO Data61, and the Motorola Solutions Foundation.

  Industry Experience

• May–Nov 2022DevOps Engineer, FPT Software, FPT Corporation
• Apr–Jul 2020Cloud Computing Intern, Bizfly Cloud Department, VCCorp Corporation
• Oct 2019–Mar 2020Cloud Computing Intern, HADTech Joint Stock Company

  Tech Stack

I enjoy modelling, formulating, and turning academic ideas into practical products.

  New blog post: Linh Nguyen, "How Transformer LLMs Work" (v1. September 2026)
  This is a living post: I will keep adding interview questions and answers on Transformer LLMs.

  News

• [10/2026] My work Not All Retained Clients Are Equal: Client Selection for Federated Unlearning has been accepted at the NewInML Workshop @ NeurIPS 2026, Paris, France. We take a new, path-dependent view of client selection: which clients are selected for unlearning depends on when and how they contributed during training.

• [09/2026] I was invited to give a talk, Data Stays, Intelligence Unites: Designing Sustainable Incentives for Privacy-Preserving AI, at the ENSURE-6G Workshop on Privacy-Preserving Machine Learning (PPML). The talk discussed why sustainable incentives, rather than algorithms alone, determine whether privacy-preserving AI systems such as Federated Learning are adopted and sustained in practice.

• [07-09/2026] I have started my research secondment at Montimage, Paris, where I will develop robust and efficient unlearning solutions for FL-based applications in next-generation networks. The work I co-authored, ‘XAI-Driven Runtime Validation for Zero-Touch Security Management in AI-Native 6G’, was accepted at NextGCom’2026.

• [08/2026] I participated in From Zero to Hero: Federated AI in Healthcare Systems Workshop and Collaborative Agent Hackathon, organized in Cambridge, UK. It was a brilliant opportunity to discuss the bottlenecks of deploying Federated Learning in practice, from regulatory to technical constraints. We also developed Flower Seed, a federated AI agent app built on Flower SuperGrid that improves post-operative outcomes for cataract surgery through federated insights across multiple facilities, without sharing patient records. You can read my reflections on the current state and future of FL here.

• [07/2026] My work Carpe Diem: Critical Learning Period-Aware Contract-Based Incentives for Federated Learning on Incentivization for Federated Learning utilizing the concept of neural network plasticity (critical learning periods), contract theory, and blockchain has been accepted for publication in IEEE Transactions on Network Science and Engineering [IF: 7.3]. In this work, we addressed the challenges of information asymmetry between clients and service provider, and assumptions of temporal homogeneity for incentivization by transforming the dynamics of the training and incentivization processes, while improving the final shared global model performance for downstream tasks. Specifically, our proposed incentive attracts clients to join and contribute resources (i.e., data) in the early training phase under constraints of budget and strategic behaviors of clients (e.g., Individual rationality, Incentive compatibility).

• [04/2026] My joint work CoCoGen+ on Coopetition (see Definition) and Federated Learning with University of Liverpool is available on arXiv and featured on a LinkedIn post by Flowers. This work models the joint effects of inter-organizational competition and data heterogeneity on organizational decision making process and social welfare. With a payoff redistribution-based incentive mechanism, CoCoGen+ mitigates competitive externalities to sustain long-term collaborations between strategic and competitive organizations.

• [02/2026] I attended the Research Ireland ADVANCE CRT Research Colloquium for specialized training on developing competitive Horizon Europe proposals. You can read my key takeaways and insights here.

• [01/2026] Our work 6G AI Security: From Fundamentals to Offensive and Defensive Landscape in 6G with multiple partners has been accepted in IEEE Communications Surveys & Tutorials [IF: 50.6].

• [01/2026] veriFUL has been accepted in IEEE Internet Computing [IF: 3.3].

• [10/2025] My joint work on Verifiable Federated Unlearning (FUL) with TU Delft is available on arXiv. This work recasts the FUL paradigm as verifiable FUL, grounded by the trust-by-design principle. Particularly, we presents a trust-by-design reference framework for verifiable federated unlearning, called veriFUL, that upholds both the right to be forgotten and the right to verify for clients (i.e., data owners) thereby strengthening client trust and privacy control in modern AI systems.

• [07/2025] My work on Federated Learning and Coopetition got accepted in the flagship IEEE Global Communications Conference 2025, Taipei, Taiwan.

• [06-09/2025] I have started my research secondment at TU Delft, Netherlands, the European Union’s Horizon Europe MSCA Staff Exchange - ENSURE-6G project.

• [05/2025] I attended 2025 CONNECT plenary to showcase my research, help in Trinity College Dublin, The University of Dublin, Ireland. You can find my poster here.

• [03/2025] I got my first research paper on Critical Learning periods and Federated Learning accepted in a semi-annual flagship 2025 IEEE 101st Vehicular Technology Conference: VTC2025-Spring, help in Oslo, Norway.

• [03/2025] I got my first SoK article on Blockchain, Federated Learning, and Data Sharing accepted at a prestigious journal ACM Computing Surveys [IF: 30.4].

• [11/2024] My team secured 3rd place in the Balbriggan Innovation Challenge - The EcoCore Hackathon (Here is Media news). We conceptualized and pitched “BART,” a transformative model aimed at overhauling transport systems in and around Balbriggan while fostering a circular economy.

• [09/2023] I started studying PhD in Computer Science at Trinity College Dublin, The University of Dublin in Ireland, fully funded by School of Computer Science and Statistics, Trinity College Dublin and Science Foundation Ireland.

• [07/2023] I just got the acceptance to attend the 6th DTU Summer Research School, which will be organized by Prof. Duong Quang Trung and Duy Tan University from July 7–9, 2023, in Da Nang City, Vietnam.

• [08/2022] I just got the acceptance to attend the 9th Vietnam Summer School of Science (VSSS09), which was sponsored and organised by the Recontres du Vietnam Foundation, ICISE and Vietnam Institute for Advanced Study in Mathematics (VIASM). Out of 700 applicants, 150 Vietnamese students and scientists were chosen for this cohort. I am so proud to be here with warm-hearted and talented people! You can find some images of mine in VSSS09 here!

  Viewers