--- schema: 1 type: video title: >- Stop Renting Intelligence: The Train-to-Deploy Loop for Specialized AI — Fireworks AI description: >- Closed models are getting expensive fast. Open models can match them, if you keep training them on your own data. Jetashree Ravi, who leads part of the applied machine learning team at Fireworks AI, explains how teams move from closed models to open ones without losing quality. Her three truths: open models like GLM 5.2 are production-ready, customization closes the remaining quality gap, and the AI stack is shifting to a continuous train-and-serve loop. She shares Fireworks' scale (about 30 trillion tokens, 10,000+ companies) and customer results: Cursor's Composer 2 trained and hosted on Fireworks with 13x faster inference, a 40x latency improvement for Vercel, and 12% better quality for Genspark with reinforcement fine-tuning. She covers the FireAttention engine and its custom CUDA kernels, speculative decoding with custom draft models, and three ways to train (a training agent, managed training, and a training API for SFT, RFT and DPO), then the flywheel: one-click deploys, monitoring, hot-loading new weights with no downtime, and matching numerics between training and serving. She ends with demos of the training agent and managed training. In this talk: • Why open models plus customization can match closed ones • FireAttention, custom kernels and speculative decoding • Training agent, managed training and the training API • Hot loading and numerical fidelity for a continuous train-and-deploy loop SPEAKER Jetashree Ravi, Tech Lead Manager, Applied ML, Fireworks AI LinkedIn: https://www.linkedin.com/in/jetashree-ravi/ GitHub: https://github.com/jraviS2023 LINKS Fireworks AI: https://fireworks.ai Managed fine-tuning docs: https://docs.fireworks.ai/fine-tuning/managed-finetuning-intro Supervised fine-tuning docs: https://docs.fireworks.ai/fine-tuning/supervised-fine-tuning CHAPTERS 0:00 Intro 1:27 Three truths about open models 2:22 Where Fireworks fits 3:12 Fireworks at scale 3:47 Cursor, Vercel and Genspark 5:27 FireAttention and custom kernels 6:27 Speculative decoding and draft models 7:22 Three ways to train 8:22 Managed training 9:27 Training API 10:07 SFT, RFT and DPO 11:07 The data flywheel 12:42 Hot loading new weights 13:22 Numerical fidelity 14:22 Demo: training agent 16:17 Demo: managed training and deploy 18:42 Wrap-up Recorded at the AI Engineer World's Fair 2026 in San Francisco. Subscribe for more talks from the engineers building with AI. AI Engineer: https://ai.engineer YouTube: https://www.youtube.com/@aiDotEngineer X: https://x.com/aiDotEngineer LinkedIn: https://www.linkedin.com/company/aidotengineer/ #FineTuning #OpenSourceAI #AIEngineer canonical: 'https://liquidgold.cc/watch/13YZ1EB9hjU' markdown: 'https://liquidgold.cc/watch/13YZ1EB9hjU.md' llms: 'https://liquidgold.cc/watch/13YZ1EB9hjU/llms.txt' parent: 'https://liquidgold.cc/channel/aidotengineer.md' channel: aidotengineer updated: '2026-10-23T22:19:03.922Z' language: null children: [] videoId: 13YZ1EB9hjU durationSeconds: 1158 publishedAt: '2026-10-11T22:30:33.000Z' searchChunks: - ordinal: 0 text: null startSeconds: 24 endSeconds: 93 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=24' heldReason: rights - ordinal: 1 text: null startSeconds: 24 endSeconds: 93 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=24' heldReason: rights - ordinal: 2 text: null startSeconds: 93 endSeconds: 104 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=93' heldReason: rights - ordinal: 3 text: null startSeconds: 104 endSeconds: 149 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=104' heldReason: rights - ordinal: 4 text: null startSeconds: 104 endSeconds: 149 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=104' heldReason: rights - ordinal: 5 text: null startSeconds: 246 endSeconds: 336 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=246' heldReason: rights - ordinal: 6 text: null startSeconds: 246 endSeconds: 336 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=246' heldReason: rights - ordinal: 7 text: null startSeconds: 336 endSeconds: 347 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=336' heldReason: rights - ordinal: 8 text: null startSeconds: 347 endSeconds: 391 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=347' heldReason: rights - ordinal: 9 text: null startSeconds: 391 endSeconds: 401 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=391' heldReason: rights - ordinal: 10 text: null startSeconds: 401 endSeconds: 448 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=401' heldReason: rights - ordinal: 11 text: null startSeconds: 459 endSeconds: 504 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=459' heldReason: rights - ordinal: 12 text: null startSeconds: 504 endSeconds: 516 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=504' heldReason: rights - ordinal: 13 text: null startSeconds: 516 endSeconds: 572 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=516' heldReason: rights - ordinal: 14 text: null startSeconds: 583 endSeconds: 607 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=583' heldReason: rights - ordinal: 15 text: null startSeconds: 607 endSeconds: 618 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=607' heldReason: rights - ordinal: 16 text: null startSeconds: 618 endSeconds: 675 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=618' heldReason: rights - ordinal: 17 text: null startSeconds: 675 endSeconds: 687 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=675' heldReason: rights - ordinal: 18 text: null startSeconds: 779 endSeconds: 802 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=779' heldReason: rights - ordinal: 19 text: null startSeconds: 815 endSeconds: 868 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=815' heldReason: rights - ordinal: 20 text: null startSeconds: 992 endSeconds: 1126 watch: 'https://liquidgold.cc/watch/13YZ1EB9hjU?t=992' heldReason: rights --- # Stop Renting Intelligence: The Train-to-Deploy Loop for Specialized AI — Fireworks AI Closed models are getting expensive fast. Open models can match them, if you keep training them on your own data. Jetashree Ravi, who leads part of the applied machine learning team at Fireworks AI, explains how teams move from closed models to open ones without losing quality. Her three truths: open models like GLM 5.2 are production-ready, customization closes the remaining quality gap, and the AI stack is shifting to a continuous train-and-serve loop. She shares Fireworks' scale (about 30 trillion tokens, 10,000+ companies) and customer results: Cursor's Composer 2 trained and hosted on Fireworks with 13x faster inference, a 40x latency improvement for Vercel, and 12% better quality for Genspark with reinforcement fine-tuning. She covers the FireAttention engine and its custom CUDA kernels, speculative decoding with custom draft models, and three ways to train (a training agent, managed training, and a training API for SFT, RFT and DPO), then the flywheel: one-click deploys, monitoring, hot-loading new weights with no downtime, and matching numerics between training and serving. She ends with demos of the training agent and managed training. In this talk: • Why open models plus customization can match closed ones • FireAttention, custom kernels and speculative decoding • Training agent, managed training and the training API • Hot loading and numerical fidelity for a continuous train-and-deploy loop SPEAKER Jetashree Ravi, Tech Lead Manager, Applied ML, Fireworks AI LinkedIn: https://www.linkedin.com/in/jetashree-ravi/ GitHub: https://github.com/jraviS2023 LINKS Fireworks AI: https://fireworks.ai Managed fine-tuning docs: https://docs.fireworks.ai/fine-tuning/managed-finetuning-intro Supervised fine-tuning docs: https://docs.fireworks.ai/fine-tuning/supervised-fine-tuning CHAPTERS 0:00 Intro 1:27 Three truths about open models 2:22 Where Fireworks fits 3:12 Fireworks at scale 3:47 Cursor, Vercel and Genspark 5:27 FireAttention and custom kernels 6:27 Speculative decoding and draft models 7:22 Three ways to train 8:22 Managed training 9:27 Training API 10:07 SFT, RFT and DPO 11:07 The data flywheel 12:42 Hot loading new weights 13:22 Numerical fidelity 14:22 Demo: training agent 16:17 Demo: managed training and deploy 18:42 Wrap-up Recorded at the AI Engineer World's Fair 2026 in San Francisco. Subscribe for more talks from the engineers building with AI. AI Engineer: https://ai.engineer YouTube: https://www.youtube.com/@aiDotEngineer X: https://x.com/aiDotEngineer LinkedIn: https://www.linkedin.com/company/aidotengineer/ #FineTuning #OpenSourceAI #AIEngineer Canonical: https://liquidgold.cc/watch/13YZ1EB9hjU · 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.