--- schema: 1 type: video title: >- Lessons from Trillion Token Deployments at Fortune 500s — Alessandro Cappelli, Adaptive ML description: >- 95% of GenAI pilots fail to reach production. Alessandro Cappelli's argument is that this isn't a deployment problem or a prompt engineering problem — it's a feedback integration problem. Instruction fine-tuning and proprietary models give you a demo. Only reinforcement learning gives you a systematic way to incorporate defects, business metrics, and production signals and keep improving. This talk covers what a production-grade RL pipeline looks like at Fortune 500 scale: synthetic data as a byproduct of environment training rather than a prerequisite, mock environments where agents can fail safely before touching real systems, and LLM judges that replace expensive annotation campaigns with a rubric-definition exercise that takes hours rather than weeks. The throughline is that agents raise the stakes on all of this — more tokens, less tolerance for errors, direct access to live databases — and RL was designed for exactly that problem. Speaker info: - https://www.linkedin.com/in/alessandro-cappelli-aa8060172 --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/X6NShR2ccOg-scaling-reinforcement-learning-lessons-from-trillion-token-deployments-at-fortune-500s Join our upcoming AI Engineer conferences (see https://ai.engineer landing page): - AIEi Shanghai, November 5–6 — the first for meeting top Chinese labs and engineers - AIE CODE in SF, November 10–12 — the top AI Coding conference returns - AIEi Sydney, December 7–8 — the first AIEi run alongside NeurIPS 2026 in Sydney --- canonical: 'https://liquidgold.cc/watch/X6NShR2ccOg' markdown: 'https://liquidgold.cc/watch/X6NShR2ccOg.md' llms: 'https://liquidgold.cc/watch/X6NShR2ccOg/llms.txt' parent: 'https://liquidgold.cc/channel/aidotengineer.md' channel: aidotengineer updated: '2026-10-24T14:11:23.090Z' language: null children: [] videoId: X6NShR2ccOg durationSeconds: 1115 publishedAt: '2026-05-12T17:00:06.000Z' searchChunks: - ordinal: 0 text: null startSeconds: 26 endSeconds: 135 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=26' heldReason: rights - ordinal: 1 text: null startSeconds: 135 endSeconds: 145 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=135' heldReason: rights - ordinal: 2 text: null startSeconds: 145 endSeconds: 237 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=145' heldReason: rights - ordinal: 3 text: null startSeconds: 248 endSeconds: 362 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=248' heldReason: rights - ordinal: 4 text: null startSeconds: 248 endSeconds: 362 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=248' heldReason: rights - ordinal: 5 text: null startSeconds: 362 endSeconds: 372 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=362' heldReason: rights - ordinal: 6 text: null startSeconds: 372 endSeconds: 464 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=372' heldReason: rights - ordinal: 7 text: null startSeconds: 464 endSeconds: 476 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=464' heldReason: rights - ordinal: 8 text: null startSeconds: 476 endSeconds: 573 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=476' heldReason: rights - ordinal: 9 text: null startSeconds: 476 endSeconds: 573 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=476' heldReason: rights - ordinal: 10 text: null startSeconds: 585 endSeconds: 714 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=585' heldReason: rights - ordinal: 11 text: null startSeconds: 585 endSeconds: 714 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=585' heldReason: rights - ordinal: 12 text: null startSeconds: 714 endSeconds: 725 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=714' heldReason: rights - ordinal: 13 text: null startSeconds: 725 endSeconds: 816 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=725' heldReason: rights - ordinal: 14 text: null startSeconds: 816 endSeconds: 828 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=816' heldReason: rights - ordinal: 15 text: null startSeconds: 828 endSeconds: 930 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=828' heldReason: rights - ordinal: 16 text: null startSeconds: 828 endSeconds: 930 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=828' heldReason: rights - ordinal: 17 text: null startSeconds: 1020 endSeconds: 1031 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=1020' heldReason: rights - ordinal: 18 text: null startSeconds: 1031 endSeconds: 1115 watch: 'https://liquidgold.cc/watch/X6NShR2ccOg?t=1031' heldReason: rights --- # Lessons from Trillion Token Deployments at Fortune 500s — Alessandro Cappelli, Adaptive ML 95% of GenAI pilots fail to reach production. Alessandro Cappelli's argument is that this isn't a deployment problem or a prompt engineering problem — it's a feedback integration problem. Instruction fine-tuning and proprietary models give you a demo. Only reinforcement learning gives you a systematic way to incorporate defects, business metrics, and production signals and keep improving. This talk covers what a production-grade RL pipeline looks like at Fortune 500 scale: synthetic data as a byproduct of environment training rather than a prerequisite, mock environments where agents can fail safely before touching real systems, and LLM judges that replace expensive annotation campaigns with a rubric-definition exercise that takes hours rather than weeks. The throughline is that agents raise the stakes on all of this — more tokens, less tolerance for errors, direct access to live databases — and RL was designed for exactly that problem. Speaker info: - https://www.linkedin.com/in/alessandro-cappelli-aa8060172 --- Watch along with the corrected transcript, summary, timestamps and resources on this talk's official AIE page: https://ai.engineer/talks/X6NShR2ccOg-scaling-reinforcement-learning-lessons-from-trillion-token-deployments-at-fortune-500s Join our upcoming AI Engineer conferences (see https://ai.engineer landing page): - AIEi Shanghai, November 5–6 — the first for meeting top Chinese labs and engineers - AIE CODE in SF, November 10–12 — the top AI Coding conference returns - AIEi Sydney, December 7–8 — the first AIEi run alongside NeurIPS 2026 in Sydney --- Canonical: https://liquidgold.cc/watch/X6NShR2ccOg · Channel: [AI Engineer](https://liquidgold.cc/channel/aidotengineer.md) > Search chunk text is held until public-source provenance is verified. ## Document chunks > Missing: no stored document chunks are available.