About Me

London, United Kingdom

I'm an AI researcher and engineer with over 7 years of experience building intelligent systems. Currently pursuing my MSc in AI at the University of Surrey, researching federated learning for privacy-preserving medical diagnosis.

At Cinnamon AI, I spent most of my time advancing deep learning research and building robust MLOps pipelines. I love the challenge of taking a research prototype and turning it into something that works reliably in production — that's where the real magic happens.

I also have a thing for cloud infrastructure. Earned my AWS Solutions Architect Professional certification on the first attempt, and I've saved companies thousands in cloud costs by actually understanding what those services do.

When I'm not writing code, you'll find me reading, strategizing in Age of Empires, or exploring London. I believe the best engineers never stop being curious.

What I Work With

Technologies and tools I use regularly

🧠

AI & Deep Learning

PyTorchComputer VisionNLP / LLMsTransformersDiffusion ModelsFederated LearningOCRMedical AI
⚙️

MLOps & Engineering

DockerKubeflowCI/CD (GitHub Actions)Neptune.aiWeights & BiasesHuggingFace HubModel Deployment
☁️

Cloud Platforms

AWS (Certified Pro)AWS SageMakerGoogle Cloud (GCP)AzureCloudFormation / IaCCost Optimization
💻

Programming & Tools

PythonGitLinux/UnixSQLBash Scripting
👥

Leadership & Soft Skills

Team LeadershipMentoringTechnical WritingAgile/ScrumCross-functional Collaboration

Certifications

🏆

AWS Certified Solutions Architect - Professional

Amazon Web Services · Sep 2023

One of the most challenging cloud certifications, earned on first attempt

Languages

Vietnamese: NativeEnglish: Fluent (IELTS 7.5)

Experience

My professional journey so far

Cinnamon AI

Full-time · 7 yrs 2 mos

Hybrid

AI Tech Lead - OCR & MLOps Focus (70%)

Oct 2021 - Present · 4 yrs 2 mos

  • Reduced model deployment effort by 31.78% and saved $6,000+/month in cloud training costs by leading a team of 3 to standardise AI models into production-ready libraries
  • Modernised document analysis pipeline with LLM-based information extraction using Azure Document Intelligence, Milvus VectorDB, and OpenAI API
  • Directed a cross-functional team of 7 to develop Kustomizable Flax, enabling scalable customizable AI training pipelines
  • Mentored 4+ junior AI engineers, enabling them to deploy 3 B2B client projects within 6 months
  • Hosted Cinnamon AI Bootcamp 2023 and 2024, evaluating 250+ applicants and teaching 20+ students annually
AIComputer VisionDeep LearningMLOpsTeam Leadership

Cloud Team Lead (30%)

Apr 2024 - Present · 1 yr 8 mos

  • Lead a team of 3 to administer IAM, established cloud governance policies, centralized logging and workflow automation
  • Drove cloud cost optimization resulting in sustained reduction of $8,200 per month in company-wide cloud expenses
  • Defined and implemented team OKRs aligned with company vision and strategy
AWSCloud ArchitectureCost OptimizationIAMTeam Leadership

Cloud Organization Administrator

Jun 2021 - Apr 2024 · 2 yrs 11 mos

  • Provided cloud technical support, permission debugging and documentation for 80+ cloud users
  • Managed and optimized infrastructure resources for AI model training across AWS SageMaker, GCE, and Ubuntu servers
  • Became the first point of contact for all cloud-related issues including RBAC, quota, deployment, and monitoring
AWS SageMakerGCPUnixCloud Administration

AI Research Engineer

Oct 2018 - Oct 2021 · 3 yrs 1 mo

  • Developed LODENet, a novel OCR model achieving SOTA accuracy and outperforming prior models by 5%; published in ICPR 2020
  • Filed a patent at Japan Patent Office (2021)
  • Led and delivered 6+ production-level B2B client projects, earning 4 "project hero" awards
  • Introduced Neptune.ai as an experiment management tool to modernize the research workflow
OCRDeep LearningResearchPatent FilingPyTorch

Education

Academic background

University of Surrey

Master of Science - MSc, Artificial Intelligence

Sep 2024 - Sep 2025 · Grade: 78/100, Distinction
  • Top-1 research institution for Computer Vision in the UK (CSRankings)
  • Dissertation: Federated Learning for Privacy-Preserving Alzheimer's Disease Classification
  • Supervised by Prof. Gustavo Carneiro

University of Information Technology

Bachelor of Science (Honours Program), Computer Science

2014 - 2018 · Grade: 8.94/10 (ranked 2nd in class of 2018)
  • Honours Program: Selective academic program with full-tuition scholarship
  • UIT Scholarships for Academic Excellence: 2015, 2016, 2017, 2018

Publications

Enhancing Diffusion Face Generation with Contrastive Embeddings and SegFormer Guidance

arXiv preprint · Aug 14, 2025

LODENet: A Holistic Approach to Offline Handwritten Chinese and Japanese Text Line Recognition

ICPR 2020 · Jan 12, 2021

Patents

TEXT LINE RECOGNITION APPARATUS, TEXT LINE RECOGNITION METHOD AND PROGRAM

Japan Patent Office · Filed Apr 6, 2021

Kind Words

What colleagues have said about working with me

"Tin is highly autonomous, resilient in pursuing targets, and impressively organized in both task execution and documentation. His professional approach and consistent reliability make him an essential part of any team."

K

Nguyen Trung-Kien

Senior AI Manager @ Cinnamon

"One of Tin's biggest strengths is strategic thinking. He doesn't just solve problems, he thinks ahead, spots potential risks early, and comes up with sustainable solutions that actually work in the long run. His ability to bridge AI and cloud made a big difference to our team's overall stability and growth."

C

Chening Yang

Lead LLM Solution Architect | Ex-AI Director @ Cinnamon

"Tin consistently delivered high-quality products through strong ownership, careful planning, and effective collaboration between technical and business teams. He stays current with the latest technological advancements and thoughtfully applies them to improve both internal systems and client-facing solutions."

D

Duc, Phan Van Hoai

Senior ML Engineer @ WorldQuant | Ex-Cinnamon

Want to know more?

Chat with my AI assistant or get in touch directly.