Elvis Saravia

Profile

Elvis Saravia is the co-founder and lead AI researcher at DAIR.AI, and one of the reasons a generation of developers can actually talk to large language models. If you have ever pasted a “few-shot” example into a prompt, added “let’s think step by step,” or copied a chain-of-thought pattern into your code, you have almost certainly been downstream of his work. His Prompt Engineering Guide — an open-source resource with roughly 70,000 GitHub stars — is the closest thing the field has to a canonical, free textbook on getting useful behavior out of LLMs. It gets cited in tutorials, corporate training decks, and university course notes, often by people who have no idea who wrote it. That anonymity is a kind of compliment: the guide has become infrastructure.

Saravia’s path to prompting fame runs through real research, not just content. He holds a PhD focused on NLP and language models, and his 2018 EMNLP paper CARER produced the widely-used emotion dataset that still shows up in countless HuggingFace tutorials. He later joined Meta AI, where he worked alongside the FAIR, PyTorch, and Papers with Code teams and was a co-creator of Galactica, the science-focused LLM that became one of the more instructive cautionary tales in the field (more below). That combination — rigorous NLP background plus a front-row seat at a frontier lab — is what makes his teaching land. He explains techniques the way someone who has shipped them explains them, not the way a marketer repackages them.

In 2020 he founded DAIR.AI (Democratizing Artificial Intelligence Research, Education, and Technologies), and today it is the hub for everything he does: the Prompt Engineering Guide, the popular AI newsletter (formerly the ML/NLP Newsletter, with 40,000-plus subscribers and its “Top AI Papers of the Week” and “AI Agents Weekly” series), and the DAIR.AI Academy’s hands-on courses on prompting, RAG, AI agents with n8n, and Claude Code workflows. He also teaches cohort-based courses on Maven and advises companies building LLM-powered products. Notably, he runs all of this from Belize — a quiet counterpoint to the assumption that serious AI work only happens in a handful of coastal cities.

For a developer learning AI today, Saravia matters because he sits at the exact seam between research and practice. He reads the papers you don’t have time for and distills them into things you can use this afternoon. He is not a hype merchant and not an ivory-tower academic; he is the person who takes a fresh arXiv preprint on some new reasoning trick and, a week later, hands you a clean explanation and a working notebook. If you want to keep pace with LLM technique without a PhD’s worth of reading, his output is one of the highest-signal feeds available.

Key Articles & Papers

Videos

Controversies

The one real controversy attached to Saravia’s name is not personal — it’s Galactica. In November 2022, Meta AI released a public demo of the model he co-authored, pitched as a tool that could summarize papers, write scientific code, and answer research questions. Within about three days, Meta pulled the demo after researchers showed it would confidently generate plausible-sounding but fabricated or biased scientific text, including fake citations and authoritative nonsense. It became an early, widely-discussed example of the “confidently wrong” failure mode that would define the LLM era. Meta’s chief AI scientist Yann LeCun publicly defended the model and criticized the backlash, arguing the tool was misjudged. In hindsight, Galactica reads less like a scandal and more like an expensive public lesson — one that arrived just weeks before ChatGPT made the same hallucination problem everyone’s concern. Saravia’s later focus on prompting, evaluation, and grounding techniques can be read, at least partly, as a constructive response to exactly the failure modes Galactica exposed.

Spotify Podcasts

[SHARE] Build AI Agents by Elvis Saravia

Courses2day.org

2026

YouTube

2025