Tech Talks · 1:17:55

Beyond the DAG: Building Agentic Workflows That Loop, Branch, and Scale

Agentic workflows loop, branch, retry, and trigger new work instead of following a fixed DAG. Santosh Kumar Radha explains how AgentField uses typed functions, structured outputs, stopping and escalation patterns, event triggers, and controls for scaling, identity, and authorization.

Santosh Kumar Radha
Santosh Kumar Radha
Sep 17, 2026
Xin

Chapters

  1. 00:00 Welcome and session context
  2. 01:53 AgentField's AI backend model
  3. 05:05 Scaling intelligence with structure
  4. 06:56 From LLM calls to agent systems
  5. 08:11 Loops, harnesses, and graphs
  6. 10:46 Just-in-time workflow graphs
  7. 13:38 Human intervention across system levels
  8. 15:15 Models, structure, and longer tasks
  9. 16:47 Programming beyond fixed graphs
  10. 18:06 Typed functions as agent services
  11. 21:26 Human workflows as architecture patterns
  12. 25:15 Event-driven workflow triggers
  13. 25:56 Code review architecture
  14. 28:22 Sieve stopping conditions
  15. 29:29 Escalating higher-order failures
  16. 31:46 Reactive data triggers
  17. 33:04 Infrastructure for stochastic functions
  18. 34:47 Large-scale pull request review
  19. 36:21 Autonomous software factory
  20. 38:17 Engineering with autonomous systems
  21. 40:14 Agents that request human input
  22. 42:57 Rethinking organizational processes
  23. 45:30 Takeaways and open-source tools
  24. 46:46 Q&A: opening discussion
  25. 47:19 Q&A: technical and nontechnical agents
  26. 50:04 Q&A: production deployment
  27. 51:00 Q&A: production architecture
  28. 52:59 Q&A: agent identity and authorization
  29. 55:43 Q&A: decentralized identity adoption
  30. 58:04 Q&A: license and hosting
  31. 59:31 Q&A: identity system integration
  32. 01:03:07 Q&A: portable agent permissions
  33. 01:05:45 Q&A: cross-system authorization limits
  34. 01:07:30 Q&A: revoking agent permissions
  35. 01:09:15 Q&A: reusable harness templates
  36. 01:11:18 Q&A: structured harness inputs and outputs
  37. 01:12:24 Q&A: interfaces for sub-harnesses
  38. 01:15:48 Q&A: planning for uncertainty
  39. 01:17:08 Closing remarks and farewell

agentic workflows · Multi-agent systems · AI agent infrastructure · agent-orchestration · Dynamic workflows · Agent identity · Agent authorization · open source AI

Summary

Fixed workflow graphs become hard to manage when agents need to loop, branch, retry, or ask for human input. Santosh Kumar Radha shows how AgentField represents agents as typed functions with structured interfaces, allowing the workflow graph to emerge at runtime. He maps planning, implementation, review, correction, and approval into reusable patterns, then covers stopping conditions, escalation paths, and event-driven triggers. The discussion also turns to production concerns: control planes, scaling, identity, authorization, policy, revocation, and when teams should reuse an existing harness.

Watch next

More from the studio

Operationalizing LLMs: From Prototype to Production1:02:51Tech Talks

Operationalizing LLMs: From Prototype to Production

Yogiraj Awati shares how Instacart moves LLM applications beyond prototypes through retrieval, offline and online evaluation, guardrails, observability, and provider fallbacks. Two case studies cover recipe ingestion and Carebot support workflows with API-backed actions and human handoff.

Yogiraj AwatiYogiraj Awati·Sep 17, 2026
Building Trustworthy Financial AI: Governance, Bias, and Mechanistic Insights48:04Tech Talks

Building Trustworthy Financial AI: Governance, Bias, and Mechanistic Insights

Fabrizio Dimino examines why financial LLM recommendations can change when option order changes. He connects positional-bias tests and mechanistic interpretability with finance-specific red teaming, risk-sensitive scoring, model validation, and AI governance.

Fabrizio DiminoFabrizio Dimino·Sep 17, 2026
Securing Code You Didn't Write: The AI Era's New Threat Model1:05:16Tech Talks

Securing Code You Didn't Write: The AI Era's New Threat Model

Coding agents expand the software supply chain to include repositories, configuration, comments, connectors, and tool output. Rome Thorstenson examines reported attacks and outlines practical controls for permissions, secrets, prompt injection, and defense in depth.

Rome ThorstensonRome Thorstenson·Sep 17, 2026