Building an AI platform: model API calls are just 2% of the system, says Adobe's Suhas Suresha
Al_Grigor · x · 2026-09-25
DataTalks.Club podcast episode with Suhas Suresha, Senior ML Engineer at Adobe and co-author of Designing AI Systems. His core claim: the model API call is only 2% of a functioning AI system; the remaining 98% is infrastructure. Topics: SDK/API design, model service (provider abstraction, routing, fallbacks, cost tracking), data service (ingestion, vector search, hybrid retrieval), and treating evals/prompt A/B testing as a platform capability.
More from coding & agent
- Agents turn stock ESP32 kits into anything: why familiar forms guide what to build — genmon · 2026-09-25
- Claude Opus codes a 20s motion design piece in 14 minutes — no After Effects, just code — techhalla · 2026-09-25
- tool-journal: SQLite-based execution journal makes agent tool-call crashes deterministically recoverable — JeremyCMorgan · 2026-09-25
- AutoGPT in practice: automation just promoted me to the worst middle manager ever — Full_Collar9026 · 2026-09-25
- How to pick a model per step in a multi-step agent? Devs weigh routing and evals — LeviYagami · 2026-09-25
- NicheDB launches a dedicated MCP servers directory for the community — Minimum_Hour519 · 2026-09-25