Manuel Sánchez Hernández
I explore the frontiers of knowledge and technology to help people and organizations.
mansanher@gmail.com
Barcelona, Spain
For over 15 years, I have helped organizations get more value from AI by building systems, teams, and products, and shaping strategy. I’ve worked with owners, executives, and technical teams at organizations of all sizes, from startups to companies with multibillion-euro revenues.
Most recently at Adevinta, I co-built and led the central machine learning organization and co-led generative AI strategy across European marketplaces. My work combined strategy, platform building, production systems, and large-scale AI adoption, improving both customer experience and internal productivity.
Earlier in my career, I built machine learning capabilities at Schibsted, developed automatic investment strategies at Morgan Stanley in London, and led product and supply initiatives in a Procter & Gamble joint venture in Barcelona.
I also teach and facilitate AI and machine learning for executive and technical audiences. Feel free to contact me if you are interested.
This site collects my writing and selected public projects on AI, LLMs, machine learning systems, software engineering productivity, and evaluation.
latest posts
| Jul 17, 2026 | Design principles of an AI-native business |
|---|---|
| Jun 21, 2026 | A summary of design patterns to secure AI agents |
| Mar 09, 2026 | A Pragmatic Evaluation of Software Engineering AI Tooling |
| Nov 11, 2025 | Rationalizing the AI bubble |
| Jul 01, 2025 | Beyond Tokens: The Context-Window Perspective on LLMs, Memory, and Mind |
| May 25, 2025 | Launching TheorIA: A Machine-Readable Atlas of Theoretical Physics |
| Apr 13, 2025 | Datasets for advancing Theoretical Physics and AI |
| Feb 01, 2025 | Selected ideas from NeurIPS 2024 |
| Jan 03, 2025 | Opening the LLM pipeline |
| Oct 06, 2024 | The path to AGI: quantifying bottlenecks |
selected projects
AI Tooling Evaluation
Evaluation of software-engineering AI tools
Kaggle Competition: Quora Insincere Questions
NLP competition under strict Kaggle kernel constraints, which finished with a Gold Medal