AI Engineer

Louis-François Bouchard Portfolio

Co-founder and CTO, Towards AI

Co-founder and CTO of Towards AI, who left a PhD at Mila/Polytechnique Montréal to build and teach. Runs "What's AI": a newsletter, a YouTube channel (74.2K subscribers, 728 videos) and a blog on practical AI engineering such as agent memory and context engineering. Co-author of Building LLMs for Production, with a second book, AI Engineering For Production, launching October 20, 2026.

Agent EngineeringRAGNewsletterAI Education

Location

Montreal, Canada

Companies

Towards AI

Credentials

  • Graduate work in AI and computer vision at ÉTS, Mila, and Polytechnique Montréal (left the PhD path)

Skills

Agent EngineeringRAGNewsletterAI Education

Notable

  • Co-author (with Louie Peters) of Building LLMs for Production; his book page reports 10,000+ copies sold and 213 Amazon ratings
  • What's AI YouTube channel: 74.2K subscribers, 728 videos
  • Co-led the "Context Engineering in 2026: Compaction, Memory & Cost" workshop at AI Engineer World's Fair 2026
  • Co-author of AI Engineering For Production, launching October 20, 2026

What makes it work

A breakdown of the choices that make this portfolio stand out.

1

An explicit research-to-practice pivot, stated directly

Naming both the Mila PhD he left ("I wanted to build and teach faster") and his current practical focus positions the content as informed by research rigor but deliberately scoped to what ships in production — a distinction the homepage draws intentionally, not by accident.

2

A stated editorial filter, backed by measured posts

The about page names the filter: new research is read with the question of what is actually useful. The featured essays follow it. "Context Engineering in 2026" measured a dozen context-management strategies on the team's production AI tutor and found that, under prompt caching, keeping everything beat every summarization strategy.

3

A structured learning path, not just a post archive

Framing content as a guide "from beginner to AI engineer" turns a blog into a sequenced curriculum, which serves readers who want a path rather than random individual posts.

What AI Engineers can take from this

Specific, actionable tips to apply to your own portfolio — no generic advice.

  • If you've moved from research to applied work, say so explicitly — it frames your practical content as informed by rigor, not a lack of depth.

  • Give your content an explicit editorial filter (what you will and won't cover) so readers know what they're getting before they subscribe.

  • Sequence your writing into a structured path for newcomers, not just a reverse-chronological post archive.

Projects

Louis-François Bouchard publishes the experiment along with the explainer: the featured context-engineering essay tests a dozen strategies on a production tutor before it gives any advice. The projects below have the same habit of measuring first.

Researchadvanced29-minute read (Aug 18, 2026)

Context Engineering in 2026: Why We Stopped Compacting Our Agent's Context

LLM agentsPrompt cachingContext compactionLocal models

A measured comparison of about a dozen context-management strategies on Towards AI's production AI tutor, run first on cached cloud models and then on local ones. It tests whether summarizing or compacting an agent's context actually helps.

It is an engineering decision made with measurements on a real production system rather than received wisdom, which is the core habit of a senior AI engineer.

  • About a dozen strategies compared on one production workload

  • Tested on both cached cloud models and local models

  • A counterintuitive result: no compaction won on cost, latency and memory

ResearchintermediateBook (2024; revised Oct 2024)

Building LLMs for Production (book)

PromptingFine-tuningRAGLangChainLlamaIndex

A practical book, co-written with Towards AI CEO Louie Peters, on adapting foundation LLMs to specific use cases through prompting, fine-tuning and retrieval-augmented generation, with coverage of LangChain and LlamaIndex. His book page quotes Jerry Liu (LlamaIndex): "The most comprehensive textbook to date on building LLM applications."

Writing the end-to-end reference on RAG and fine-tuning requires having shipped and debugged those systems. The book is a compact record of that practice.

  • Covers prompting, fine-tuning and RAG as options to choose among, not a single recipe

  • Endorsed by LlamaIndex's Jerry Liu on the book page

  • Followed by a second book focused on harnesses, deployment, evaluation and monitoring

Ready to build your portfolio?

Follow the AI Engineer roadmap — skills, projects, and timeline to get hired.

AI Engineer Roadmap