AItomate

Hi, I’m Hoang.

I run a solo consulting practice focused on helping startups build better practices as many software companies transition to AI

I specialize in AI applications and enterprise solutions, combining practical innovation with a deep understanding of business needs.

With experience spanning startups and frontier AI labs, I’ve built systems from the ground up, scaled them, and navigated the intersections of research, product development, and psychology.

I’m confident in helping your company focus on the right AI layers to drive meaningful impact.

I’ve seen companies waste months and millions chasing the wrong priorities. While I can’t guarantee success or product-market fit, I can help you create a clear plan, guide your team through simple yet effective experiments, and focus on what truly matters.

My job is straightforward: to tell you when what seems easy is hard, and when what seems hard might actually be simple.

Success is about minimizing regret, and I’m here to help you do just that.

• Nov 2024 – Present
  Member of Technical Staff @ OpenAI

  

  I joined OpenAI – building a new enterprise product.
  
  I will update here as product progress.

• Jun 2021 – Nov 2024
  Senior Research Scientist @ Snorkel AI

  

  Snorkel AI is an AI Unicorn ($1B valuation) spin-off from Stanford AI Lab – focused on data quality and curation.
  
  Half of my time, I focused on researching LLMs – alignment, agentic workflows, and guardrails.
  
  The other half of my time, I consulted Snorkel’s customers – large enterprises on designing performant and scalable LLM Systems.

• Aug 2019 – Nov 2020
  Co-Founder @ Vizly

  VIZLY

  I co-founded Vizly – a platform to understand clients’ design preferences through their quick reactions using the 20 second gut test and AI attention prediction.
  
  We were a part of Mozilla Foundation Spring 2020 Fix-the-Internet Incubator.

• Others non-listed period

  Fujitsu AI Labs
  Developed Graph Neural Networks models and contributed to building an internal AI platform
  
  DeepCore (SoftBank’s AI Startup Fund)
  Created an AI-powered digital meeting summarizer (precursor to tools like Otter.ai), delivering notes, action items, and meeting minutes
  
  Pictory.AI
  Designed a text summarizer optimized for seamless integration with image search capabilities

  Minerva University – research with the Head of Social Sciences
  Conducted behavioral analysis research. Presented findings at the AAC&U Conference 2021.
  
  Microsoft
  Collaborated with the Microsoft Garage Director to enhance engagement strategies for specialized ERGs
  
  VNG Corp.
  Web Platform Team – analyzed user behavior to streamline workflows, reducing time-to-first-purchase
  
  Kakao Corp.
  Developed a leading platform to motivate social responsibilities for 100 days – project100.kakao.com

Notable Work

I am the lead author of the first open-source LLM employing “Iterative DPO Alignment” (2023), ranked 2nd on the AlpacaEval leaderboard in Jan 2024. I was invited by Microsoft to present this work at Nvidia GTC 2024.

My approach, involving iterative offline alignment (vs. RLHF), has inspired follow-up papers and models from FAIR, Salesforce, and UCLA, such as “Iterative Preference Learning from Human Feedback“, “Iterative Reasoning Preference Optimization,” and “Self-play Fine-tuning (SPIN) and Preference Optimization (SSPIN).“

This work is concurrent with FAIR’s “Self-Rewarding Language Models.”
* FAIR = Meta AI Research

Selected Publications, Guest Speakers, and Presentations

• [Guest Speaker, 2024] AAAI Symposium Spring @ Stanford: Invited to present research on “Clinical Foundation Models” utilizing programmatic data development for private proprietary data.
  
• [Guest Speaker, 2024] Nvidia GTC – Invited by Microsoft to present my research on programmatic alignment, iterative DPO, and RLAIF.
  
• [Open-source model and dataset, 2024] Lead author of Snorkel-Mistral-PairRM-DPO – ranked 2nd place model on the AlpacaEval LLM leaderboard Jan 2024 – Iterative DPO alignment with reward models
  
• [Open-source model & Blog Post, 2023] Collaboration with Together AI to release RedPajama-Chat-Curated. Achieved a 24% improvement by optimizing data mixtures and training with fewer data (first author).
  
• [Guest Speaker, 2023] “Fine-Tuning and Customizing LLMs for Enterprise” – LLMs Summit, Snorkel AI – (with Prof. Percy Liang – Stanford, Douwe Keila – RAG’s author, and Joe Spisak – Meta’s Llama)
  
• [Book Co-Author, 2017] “Advanced Mathematics for Undergraduate Olympiads” – Vietnam.

Other Team Members

At Aitomate, I lead a talented team of AI/ML engineers with diverse experience at top organizations across Vietnam, Singapore, Australia, and beyond. Together, we bring a deep understanding of cutting-edge technologies and practical problem-solving to deliver exceptional results.

As your main point of contact, I ensure clear communication and seamless collaboration, involving the right team members with specialized expertise as your project evolves.