Leveraging Gemini & Web-LLM for Intelligent Experiences
summary
Claims a working Angular app that uses Web-LLM (Llama 3.1 8B) downloaded and cached in ~5s to run fully offline, plus cloud fallback via Gemini 3 via API; demos include a streaming chat, an adaptive quiz and a newsletter generator assembled in about 4 hours. Reports on-device latency ~100–500 ms versus cloud 1–2 s, a cost example of ~$625/day for 100k users with cloud AI, zero per-request API cost when local, and notes about needed fixes, tests and security for production.
AI-written summary · about
chapters (11)
- 0:00AI in Angular intro
- 1:53JavaScript cost dilemma
- 6:51On-device Web LLM
- 11:05Angular with Signals
- 13:53Intelligent adaptive UIs
- 17:25AI coding quality & security
- 20:32GitHub Copilot for AI dev
- 25:09AI dev real-world results
- 28:39Hard truths on AI in dev
- 30:05AI in Angular future
- 32:56Q&A
related talks
46mFrom ML to LLM: on-device AI in the browser
16mReact + WebLLM: Building Intelligent Applications with In-Browser Language Models
40mTechniques to avoid common LLM pitfalls
34mPrivacy First: Building LLM-Powered Web Apps with client side WASM
45mAvoid common LLM pitfalls
05mBuild your own LLM on Cloud
31mFrom ML to LLM on device AI in the browser
51mGemini, Google's Large Language Model