Production readiness, architectural patterns, multi-agent interoperability, and enterprise security. Moving from single-prompt wrappers to resilient, autonomous multi-agent systems.
Engineered for systems architects, staff engineers, AI researchers, and technical leaders bridging the gap between toy prompts and production agent clusters.
Focused on the hardest infrastructure challenges when scaling agentic workloads into enterprise production environments:
Facilitated, code-first workshops where engineers build, harden, and inter-operate production-grade agents in real-time:
/.well-known/agent.json and orchestrating async state lifecycles.
Badge pickup, coffee, Wi-Fi onboarding, and A/V soundchecks across all stages.
Moving from Single-Prompt Wrappers to Autonomous Multi-Agent Systems. Transitioning to resilient, stateful agent networks, deterministic checkpoints, and production uptime engineering.
Cross-Organizational Communication & Protocols. Open standards, decentralized ecosystems, dynamic capability registries, and cross-boundary negotiation.
Orchestration vs. choreography, state management, human-in-the-loop (HITL) checkpoints, retry storms, and durable multi-day execution engines.
Cryptographic agent identities, capability registries, SPIFFE/SPIRE attestation, and verifying third-party agents prior to context handoff.
MicroVMs, Wasm & AI-DLC Isolation. Memory safety, cold-start mitigation, and hypervisor-level isolation for untrusted code execution.
Tool Definitions & Auth. Exposing enterprise resources via standard MCP interfaces, transport security (SSE/Streamable HTTP), and scoped OAuth.
Prompt Injection & Egress Filtering. Defense-in-depth against indirect injection, confused deputy scenarios, and real-time PII redaction.
Mitigating Tool Poisoning. Defending against rug-pull attacks, schema manipulation, and enforcing gateway security policies programmatically.
OpenTelemetry & Evidence Artifacts. Mapping non-deterministic execution paths into structured trace trees and verifiable audit logs.
Agent Cards & Task Lifecycles. Publishing /.well-known/agent.json, handling asynchronous task state changes, and negotiating capabilities.
Path Grading & Drift Analysis. Assessing decision-tree efficiency, detecting cyclic loops, and evaluating context-causality metrics.
Full Pipeline Integration. Connecting an orchestrator agent to remote A2A peers backed by private MCP servers using distributed trace IDs.
Both tracks reconvene for final key takeaways, feedback collection, community announcements, and informal networking.
Speakers and session leads must align their content with rigorous production depth, actionable code, and concrete architectural trade-offs.
Scope: Deterministic state machines vs. dynamic loops. Handling retry storms, infinite loops, and persistence models for multi-day runs.
Scope: Cross-organizational agent-to-agent negotiation protocols, dynamic catalogs, zero-trust perimeters, and capability matching.
Scope: User-delegated OAuth vs. machine-to-machine identities, cryptographic attestation, SPIFFE/SPIRE, and ephemeral credentials.
Scope: Docker containers vs. MicroVMs (AWS Firecracker) vs. Wasm isolation for executing untrusted agent code safely.
Scope: Centralized proxy layers, token budget management, dynamic PII masking, and semantic policy checks prior to outbound calls.
Scope: Threat modeling indirect prompt injection in RAG, confused deputy exploits, tool-poisoning, and lateral movement.
Scope: Mapping OpenTelemetry (OTel) across async decision trees. Generating cryptographically signed Evidence Artifacts for compliance.
Scope: Evaluating non-deterministic agent runs: path optimality, redundant tool calls, trajectory drift, and context causality checks.
The afternoon hands-on track provides practical, code-first workshops focused on real-world MCP and A2A implementations.
To ensure all participants are properly configured for the hands-on sessions, repository access, dependency setup instructions, and environment prerequisites will be sent directly to registered attendees based on their confirmed registrations prior to the event.
Constructing enterprise-ready MCP endpoints and defending against tool poisoning and rogue parameter injection.
Expose databases and internal ticketing endpoints via standardized MCP JSON-RPC tool declarations.
Secure transport via Server-Sent Events (SSE) and Streamable HTTP with scoped OAuth tokens.
Implement defensive schema validation to guard against rug-pull attacks designed to hijack model tool selection logic.
{
"name": "enterprise-tickets-mcp",
"version": "1.0.0",
"transport": "sse",
"auth": {
"type": "oauth2_bearer",
"scopes": ["tickets:read", "tickets:write"]
},
"guardrails": {
"sanitize_tool_descriptions": true,
"max_payload_bytes": 1048576
}
}
Implementing cross-agent communication protocols, asynchronous task states, and federated peer delegation.
Publish and parse standard capability contracts at /.well-known/agent.json.
State machines for submitted, working, input-required, completed, and failed.
Enforce cryptographic verification before wiring orchestrator agents to remote A2A peers backed by MCP servers.
{
"agent_id": "aaif.dallas.billing-agent-01",
"name": "Enterprise Billing Agent",
"protocol_version": "a2a/v1.0",
"capabilities": ["invoice.generate", "ledger.reconcile"],
"attestation": {
"spiffe_id": "spiffe://aaif.internal/agent/billing",
"public_key": "ecdsa-sha2-nistp256 AAAAE2V..."
}
}
World-class systems architects, security researchers, and lead maintainers driving autonomous agent infrastructure from concept to production.
Announcing Soon
AI Systems & Architecture
Pioneering reliable, stateful agent networks, deterministic checkpoints, and multi-day long-running workflow persistence models.
Announcing Soon
Protocols & Open Standards
Architecting decentralized capability catalogs, zero-trust perimeter federation, and inter-organizational task delegation.
Announcing Soon
MicroVMs & Runtimes
Analyzing hypervisor isolation, AWS Firecracker microVM cold-start mitigation, and memory safety during runtime code synthesis.
Announcing Soon
Gateways & Guardrails
Defense-in-depth against indirect prompt injection, confused deputy attacks, real-time PII redaction, and semantic policy proxies.
Announcing Soon
Model Context Protocol Lead
Facilitating live code construction of enterprise MCP servers with Server-Sent Events (SSE), scoped OAuth, and input validation.
Announcing Soon
Distributed Multi-Agent Systems
Guiding engineers through Agent Cards (/.well-known/agent.json), asynchronous task states, and federated peer delegation.
We welcome technical lightning talks (10 min) showcasing real-world lessons learned, sandbox benchmark data, or open-source tooling.
Supported by leading industry innovators, open-source consortiums, and engineering communities committed to advancing autonomous AI systems.
Advancing technology for humanity through joint community technical workshops, student mentorship, and engineering symposia.
Global non-profit foundation dedicated to open-source agentic AI protocols, decentralized multi-agent interoperability standards (MCP & A2A), and production readiness benchmarks.
University of Texas at Dallas premier student AI organization driving machine learning research, agentic hackathons, and projects.
UT Dallas IEEE Student Branch empowering next-generation engineers through technical workshops and community engagement.
Southern Methodist University IEEE Student Branch fostering innovation, technical mentorship, and engineering excellence.
Connect directly with 400+ systems architects, engineering leaders, and practitioners deploying production agent networks. Collaborative sponsorships include Food & Hospitality (Breakfast, Lunch, Snacks), official AAIF Member Products & Swag (merch.aaif.io), and Hands-On Lab Infrastructure.