Production ML Deployment Architecture & System Diagnostics
Design production ML systems, diagnose failures, optimize architecture
What You Can Do
This skill helps you architect production-grade ML deployments, diagnose failures in distributed systems, and optimize performance bottlenecks through systematic architectural analysis. You'll design resilient inference pipelines, troubleshoot data flow issues, and identify optimization opportunities using proven MLOps patterns. Get actionable recommendations grounded in architecture-first thinking rather than point fixes.
Features
Build containerized inference pipelines with load balancing, auto-scaling, and multi-region strategies
Systematically trace failures across distributed components (data ingestion, model serving, orchestration)
Analyze latency, throughput, and resource utilization across GPU/CPU/memory/networking
Design redundancy, circuit breakers, fallback strategies, and graceful degradation patterns
Set up metrics, traces, logs, and alerting hierarchies for production systems
Right-size infrastructure, identify wasteful patterns, and balance performance vs. cloud spend
Design data validation, feature engineering workflows, and data quality monitoring
Example Output
Example 1: Deployment Architecture for Real-Time Inference
Component Topology:
┌─────────────────┐
│ Load Balancer │ (health check every 5s)
└────────┬────────┘
│
┌────┴────┐
│ │
┌───▼──┐ ┌───▼──┐
│ Pod 1 │ │ Pod 2 │ (auto-scale 2-10 pods based on QPS)
└───┬──┘ └───┬──┘
│ │
└────┬─────┘
│
┌────▼──────────┐
│ Feature Store │ (cached, 500ms TTL)
└────┬──────────┘
│
┌────▼──────────┐
│ Model Server │ (TorchServe + GPU)
└────────────────┘
Key Decisions:
• p95 latency: 120ms (50ms model, 30ms network, 40ms cache)
• Failover: Circuit breaker opens after 5 errors
• Auto-scale trigger: CPU > 70% for 2 minutes
Example 2: Production Failure Diagnosis
Issue: Inference latency spiked to 2s; error rate hit 15%
Root Cause Chain:
- Feature Store cache eviction (Redis memory full)
- Cache misses → DB fallback queries
- DB bottleneck → slow feature fetches
- Model server queued requests → timeout after 30s
- Client retries → cascading failures
Immediate Fix: Increased Redis memory, implemented cache prewarming Prevention: Set alerts at 80% memory; implement bulkhead isolation for feature store timeouts
Example 3: Cost Optimization Plan
Current: 4x GPU instances, $12k/month (avg utilization 25%)
Optimizations:
- Reduce to 2x GPU + spot pricing (-60% cost)
- Implement request batching (collect 32 predictions)
- Add CPU fallback for low-latency requests
- Cache 80% of prediction outputs
Result: $4.8k/month (60% savings) with p95 latency unchanged
What's Included
- SKILL.md file: Complete skill with decision trees, diagnostic workflows, and architecture templates
- Deployment Architecture Templates: Reference blueprints for single-region, multi-region, and edge ML deployments
- Failure Diagnosis Checklist: Systematic protocol to trace issues across data ingestion, model serving, and orchestration layers
- Performance Analysis Worksheet: Structured format to collect latency, throughput, and resource utilization metrics
- Monitoring Setup Guide: SLO definitions, alerting thresholds, and dashboard templates for production ML systems
- Resilience Pattern Library: Circuit breakers, bulkheads, timeouts, retry logic, and graceful degradation patterns
- Cost Analysis Template: Cloud resource costing model and optimization opportunities worksheet
Who It's For
- ML/Platform Engineers — deploying and maintaining production ML systems at scale
- DevOps/SRE teams — managing infrastructure reliability and performance optimization
- ML Research Leaders — scaling experimental models to production workloads
- Solutions Architects — designing enterprise ML deployment strategies
- System Performance Engineers — optimizing latency-sensitive ML workloads
Best For
- Architecting containerized ML inference systems at scale
- Diagnosing and resolving production outages in ML pipelines
- Optimizing GPU utilization and reducing compute costs
- Designing multi-region or edge ML deployment strategies
- Implementing monitoring and observability for production ML systems
You might also like
You get Claude to analyze your API specifications and automatically generate bulletproof parameter validation logic. Claude creates security-hardened middleware, enforces type constraints, detects malicious input patterns, and generates comprehensive error handling — all tailored to your framework and existing codebase. The result is production-ready validation code with zero unvalidated inputs reaching your business logic.
This skill enables you to automatically audit your websites against compliance standards and detect deviations from expected baselines. You can track regulatory requirements, identify policy violations, and generate compliance reports with actionable remediation steps. Monitor multiple sites simultaneously and maintain detailed audit trails for compliance documentation.
Transform schema and data changes into production-ready migration packages that protect data integrity and minimize downtime. You can handle column additions, renames, type conversions, and backfills using proven patterns like expand-contract and shadow tables. The skill generates explicit rollback scripts, validation checklists, and phased execution plans that keep applications stable throughout the migration.
You can conduct comprehensive toxicology safety assessments, generate regulatory-compliant safety documentation, and produce detailed hazard analyses for pharmaceutical and chemical submissions. This skill structures toxicological data into standardized safety reports, maps findings to regulatory guidelines (FDA, EMA, ICH), and identifies risk characterization strategies to support regulatory submissions and approval timelines.
You can submit your Solidity smart contracts for comprehensive security analysis, vulnerability detection, and gas optimization recommendations. Claude reviews your code against known attack vectors, design patterns, and best practices—identifying reentrancy risks, overflow/underflow issues, access control flaws, and inefficient operations. You receive actionable fix recommendations with explanations of the security implications and gas savings potential.
You can design end-to-end NGS pipelines tailored to your experiment type (WGS, WES, RNA-seq, targeted), troubleshoot pipeline failures by analyzing QC metrics and logs, and optimize performance by fine-tuning parameters and identifying bottlenecks. This skill handles the full workflow from raw sequencing data quality assessment through variant interpretation and annotation.
This skill guides you through designing Kalman filters for your specific state estimation and sensor fusion problems. You'll learn to diagnose filter divergence, tune covariance matrices (Q and R), and validate observability—turning mathematical theory into production-ready filters. The skill provides step-by-step workflows, diagnostic tools, and practical templates for common applications like GPS/IMU fusion, SLAM, and aircraft navigation.
You can accelerate your entire antenna design workflow—from initial concept to optimized specifications—by leveraging Claude's knowledge of electromagnetic theory, antenna types, and design trade-offs. Claude helps you parameterize designs, identify optimal configurations for your frequency band and application, and prepare detailed briefs for simulation and manufacturing. You'll generate design reviews, impedance-matching solutions, and validation checklists without context-switching to multiple tools.