Why Large? Tiny LMs & Agents on Edge/Robotics
Conference Context
- Date/time: 2026-06-30 · 2:50pm-3:10pm
- Track/room: Robotics & World Models · Track 2
- Speaker(s): Cormac Brick
- Session type/status: sponsor · confirmed
- Track: Robotics & World Models
- Room: Track 2
- Session type: sponsor
- Status: confirmed
Session Description
big models get a lot of press. small model scale much better. RAM is expensive. The real world needs tiny models for scale on the edge. This workshop will cover how to combine both for mobile and robotics deployment. specifically covering: - skills are different on mobile - tiny LLMs <1B scale much further on mobile/web - how to fine tune and train tiny models. - skills on robotics / edge/ mobile - latest open models for edge (including gemma, qwen, and anything else that happens in next 10 weeks) This talk will focus on open models, including some gemma variants that will be shortly announced.
Media Evidence
From 46% to 90%: Fine-Tuning Tiny LLMs for On-Device Agents — Cormac Brick, Google (speaker-match related prior/adjacent AI Engineer video; captions: English auto-captions).
- Source video:
youtube--TiET_K-E_g - Slide deck: Dense Slides: From 46% to 90%: Fine-Tuning Tiny LLMs for On-Device Agents — Cormac Brick, Google — 1 visible slide image(s); 1 HTML recreation(s).
- Additional slide evidence: Slides: From 46% to 90%: Fine-Tuning Tiny LLMs for On-Device Agents — Cormac Brick, Google, Reconstructed Slides: From 46% to 90%: Fine-Tuning Tiny LLMs for On-Device Agents — Cormac Brick, Google
- Slide-derived themes for
youtube--TiET_K-E_g: running, edge, many, benefits, faster, network, involved, sensitive.
Evidence Graph
This evidence graph is generated from currently linked source material: official schedule text, related video pages, cached transcripts, visible slide text, dense/reconstructed slide pages, and AI slide-classification audits.
Media Signals
youtube--TiET_K-E_g— 5 slide-derived text signals- Slide-derived themes for
youtube--TiET_K-E_g: running, edge, many, benefits, faster, network, involved, sensitive. - Evidence links for
youtube--TiET_K-E_g: youtube TiET_K E_g, youtube TiET_K E_g slides, youtube TiET_K E_g dense slides, youtube TiET_K E_g reconstructed slides
Agent Reading Notes
Use these signals to refine the synopsis, topic links, people/company context, and method notes. If a source is a related external video rather than an exact official recording, keep it framed as supporting evidence.
Transcript Status
Related video transcript availability: English auto-captions. Treat this as supporting context, not a recording of this exact scheduled session unless later confirmed. Not fetched yet.
People
Supporting Slides
- youtube TiET_K E_g slides — extracted from the related public AI Engineer video.
Slide Evidence
- Slide-only cropped deck: youtube TiET_K E_g dense slides (1 viable slide images).
- Related slide/OCR pages:
- youtube TiET_K E_g dense slides
- youtube TiET_K E_g reconstructed slides
- youtube TiET_K E_g slides
- Slide-derived terms:
engineering,future,skills,tlms,engineer,edge,example,restaurant,roulette,tral,braintrust,workos,openal,tinyllmsand,agentsonedgedevices,bringing,state-of-the-artagenticskilstothe,edgewithopenmodels
Synthesis
Synthesized Breakdown
Why Large? Tiny LMs & Agents on Edge/Robotics ## Conference Context - Date/time: 2026-06-30 · 2:50pm-3:10pm - Track/room: Robotics & World Models · Track 2 - Speaker(s): Cormac Brick - Session type/status: sponsor · confirmed - Track: Robotics & World Models - Room: Track 2 - Session type: sponsor - Status: confirmed ## Session Description big models get a lot of press. small model scale much better. RAM is expensive.
Speaker And Company Context
- Cormac Brick — Principal Engineer, Google AI Edge at Google.
Topics Covered
Derived Links And Source Material
- youtube TiET_K E_g — related YouTube source page.
- youtube TiET_K E_g slides — slide evidence.
- youtube TiET_K E_g reconstructed slides — slide evidence.
- youtube TiET_K E_g dense slides — slide evidence.
Novel Concepts / Clever Methods
- No highlighted novel concept has been detected yet.
Evidence Boundary
This synthesis is based on the official schedule and linked source pages. It should be revisited when exact session recordings or transcript-backed secondary sources are available.