AI Engineer Summit 2023: Deployment, Compatibility, Multimodality
Eugene Yan's reflections from AI Engineer Summit 2023 highlight the practical hurdles teams face when shipping AI systems.
AI & ML
Frontier models, research results, agents, and developer tooling.
Four Mechanisms That Keep Machine Learning Projects on Track
A practical guide from Eugene Yan on using pilots, reviews, and timeboxing to reduce risk in ML initiatives.
Essay: AI X-Risk Coverage Is Credulous and Cartoonish
A new essay argues that mainstream reporting on AI existential risk lacks rigor and borrows from children's cartoons.
When Dependency Teams Can't Help: A Practical Guide
Eugene Yan offers strategies for engineers when upstream teams are unavailable, emphasizing understanding, trust, and preparation.
A Community-Built List of Commercially Usable LLMs
Eugene Yan's open-LLMs list fills a gap by tracking models with permissive licenses, refined through community pull requests.
Goodbye Roam, Hello Obsidian: A Notes Migration Guide
A practical walkthrough of moving a personal knowledge base from Roam Research to Obsidian, including cross-device sync for notes and images.
RecSys keynote and promotion capped Eugene Yan's 2022
A senior machine learning engineer's year-end review highlights career growth, conference recognition, and community building in applied ML.
Data Labeling Guidelines: A Key Lever for Reliable ML
For AI teams, clear annotation instructions can reduce labeling errors and speed up data production.
Simplicity Wins in Practice, but Complexity Still Sells
An essay by Eugene Yan argues that the technology industry rewards elaborate designs even when simpler ones would work better.
RecSys 2022 Keynote Asks: Is the Juice Worth the Squeeze?
A keynote at RecSys 2022 challenges recommender systems teams to weigh the real-world costs of model complexity against the benefits they deliver.
A Practical Onboarding Playbook for New Tech Hires
Eugene Yan's guide distills onboarding into a 100-day plan that balances learning with early action to build trust.
Using Randomness to Counter Position Bias in Recommender Systems
A practical guide says rankers can fight position bias by adding randomness or learning from the randomness that already exists in user behavior.
High-Level Intent vs Low-Level Requirements in AI Workflows
Eugene Yan breaks down when AI systems should be told what to achieve versus exactly how to do it.
Why Robust Data Pipeline Tests Mean Fewer Integration Tests
A new essay argues that data and ML pipeline reliability comes from better unit tests, not more integration tests.
LLM Experiments for Research and Planning Hit Retrieval Limits
Eugene Yan's experiments show LLMs can assist with research, reflection, and planning, but document retrieval remains a weak point that search and recommender methods can improve.
Recognizing Design Patterns in Machine Learning Systems
A new guide helps practitioners spot recurring structural patterns in ML code and infrastructure so components are used as intended.
Red Flags to Watch For Before Joining a Data Team
A practical checklist for spotting warning signs in data roles, from data quality to management and tooling.
RecSys 2022 Recap: Favorite Papers and Practical Lessons
A single practitioner's recap of RecSys 2022 highlights three standout papers, 17 summaries, and lessons that go beyond machine learning.