SessionExpo trackconfirmed

Your Stack Has a Latency Problem You Can’t See

Day
Day 3 — Session Day 2 · Wed, Jul 1
Time
2:25pm-2:45pm
Room
Expo Stage 4 SE
Track
Share
Track theme(approximate)
Agent Architecture & Multimodal Pipelines (Expo)

Sponsor talks on agent architecture decisions (prompt/memory/weights), multimodal models, FDE support agents, and SFT data pipelines.

Accessible with the Expo Explorer pass and above.

About this session

Break down a real AI voice call path step by step. Show where time actually goes: network hops between providers, handoff latency, buffering, connection overhead. The model is rarely the bottleneck. The gaps between vendors are. What changes when inference, STT, TTS, and telephony run on co-located infrastructure. One network, zero inter-provider hops. Show the before/after latency breakdown. Zoom out to the inference economics. Owned GPUs, not rented. FP8 throughput on FOSS models. Pricing that follows the cost of compute, not cloud provider markup. The voice use case is the proof. The infrastructure story is the point.

Speakers

To be announced.