Workshop Details
Data Discovery Workshop Series
Engineering Beyond the “Rule of Eight”
Building Resilient, Production-Grade LLM Systems Through Functional Factoring & Agent Orchestration
Speaker
Date
June 13, 2026; 4:30 PM
Venue
In-Person : Mercer Island Library Hall 1
4400 88th Ave SE, Mercer Island, WA 98040
Virtual: Microsoft Teams
4400 88th Ave SE, Mercer Island, WA 98040
Virtual: Microsoft Teams
Duration
45 Minutes
Audience
AI Engineers, ML Architects & Platform Teams
Registration
Workshop Content
Engineering Beyond the “Rule of Eight”
Enterprise Architecture for Distributed Agentic Workflows: Governance, Sovereignty, Resilience & Optimization
Architecture
Multi-Agent
Swarm Orchestration
Governance
Policy-as-Code
Autonomous Loops
Compliance
EU AI Act
Data Sovereignty
Scale
Petabyte
Concurrent Execution
1
Deconstructing the “Rule of Eight” Bottleneck
Identifying Architectural Flashpoints When Scaling Beyond Single-Digit Micro-Orchestrations
Learning Objectives
- Identify and measure the exact architectural flashpoints that emerge when scaling multi-agent AI workflows beyond eight concurrent agents
- Quantify cross-agent synchronization overhead and its exponential growth patterns across distributed orchestration topologies
- Diagnose context window degradation as shared state expands across distributed agent pools
- Model network latency amplification in multi-hop agentic communication chains and its compound effect on end-to-end throughput
- Apply the “Rule of Eight” diagnostic framework to predict scaling ceilings before production deployment
2
Architecting Autonomous Governance Loops
Real-Time Policy-as-Code Guardrails in Distributed Data Fabrics
Learning Objectives
- Design real-time, policy-as-code guardrails that enforce corporate, safety, and operational boundaries dynamically
- Embed governance enforcement directly into distributed data fabrics without introducing manual human-in-the-loop dependencies
- Implement self-healing policy enforcement that adapts to evolving regulatory and operational requirements in real-time
- Construct autonomous feedback loops where governance violations trigger automatic remediation rather than workflow halts
- Balance agent autonomy with organizational accountability using tiered escalation and risk-classification frameworks
3
Enforcing Hard Geographic Data Sovereignty
Boundary-Aware Metadata Tagging & Dynamic Data Routing for Global Compliance
Learning Objectives
- Implement boundary-aware metadata tagging that programmatically classifies data by jurisdictional sensitivity at the point of creation
- Design dynamic data routing patterns that restrict sensitive payloads to compliant geographic regions without manual intervention
- Ensure absolute alignment with complex global compliance frameworks including the EU AI Act, EU Data Boundary, GDPR Articles 44–49, and China PIPL
- Build routing decision engines that adapt in real-time as regulatory landscapes shift — incorporating new adequacy decisions and treaty changes automatically
- Prevent sovereignty violations at the infrastructure layer, making non-compliance architecturally impossible rather than merely policy-prohibited
4
Mitigating Cascade Failures in Agentic Swarms
Predictive Error-Handling & Telemetry for Rogue Agent Detection and Neutralization
Learning Objectives
- Formulate predictive error-handling strategies that detect misaligned or rogue agent behavior before it cascades downstream
- Design telemetry architectures that isolate malfunctioning agents without disrupting healthy swarm operations
- Neutralize infinite-loop compute spikes through resource boundary enforcement, circuit breakers, and automated kill switches
- Prevent downstream data corruption from partially-completed or poisoned agent transactions using transactional isolation buffers
- Implement swarm-wide health scoring that predicts cascade failure probability in real-time and triggers preemptive intervention
5
Optimizing Infrastructure for Continuous Training
Data-Shaping, State Management & Structural Integrity at Petabyte Scale
Learning Objectives
- Apply advanced data-shaping principles at the bedrock engineering layer to maintain structural integrity during continuous model training
- Prevent model poisoning through schema-on-write validation, distribution drift detection, and provenance-weighted training pipelines
- Minimize memory footprints during petabyte-scale concurrent model execution using gradient checkpointing and mixed-precision strategies
- Design state-management architectures that support rolling model updates without downtime or prediction inconsistency
- Implement validation gates and canary deployment patterns that detect training data corruption before it influences production model weights
Facilitator: Harsh Ranjan
Senior Software Engineer at Snowflake
Senior Software Engineer with experience building large-scale backend and data platforms across Snowflake, Microsoft, and Facebook. Currently at Snowflake, working on data governance and data quality foundations that power monitoring, profiling, and trust at scale. Design and ship production-grade backend systems, including data quality expectations, metric-driven frameworks, privilege-aware APIs, and system functions used by both UI and programmatic consumers. Previously at Microsoft and Facebook, I worked on distributed systems and telemetry-driven services, gaining deep experience in scalability, on-call ownership, and cross-team execution in high-impact environments. I’m comfortable operating in ambiguous problem spaces, mentoring engineers, and driving complex initiatives from design through rollout.
Event Throwbacks
