Writing

Technical articles on systems architecture, agent execution, and operational patterns.

· Implementation

Observability for AI Agent Systems

Why traditional APM tools fall short for agent workloads, and how to build observability that captures token usage, decision traces, and tool call latencies.

observability agents operations
· Architectural

Memory Systems for Long-Running Agents

Exploring working memory, episodic memory, and semantic memory architectures for agents that need to maintain context across hundreds of interactions.

agents memory architecture
· Implementation

Building Reliable Agent Loops: Retry, Backoff, and Circuit Breakers

How to design agent execution loops that handle failures gracefully — covering retry strategies, exponential backoff, circuit breakers, and graceful degradation patterns for production AI systems.

agents reliability systems-design
· Implementation

Designing Tool Interfaces for LLM Agents

Practical patterns for designing tool APIs that LLMs can actually use reliably — covering parameter design, error messages, and output formatting.

agents tool-use api-design
· Implementation

Structured Output Parsing in Production

Battle-tested strategies for getting reliable structured data from LLMs — covering JSON mode, schema validation, retry logic, and partial extraction.

llm parsing reliability
· Architectural

Context Window Management for Multi-Turn Agents

Strategies for managing the context window across long conversations — sliding windows, summarization, priority eviction, and hybrid approaches.

agents context-management architecture
· Advisory

Evaluating Agent Systems Beyond Accuracy

Why accuracy alone is a misleading metric for agent systems, and how to build evaluation frameworks that capture cost efficiency, latency, reliability, and user trust.

evaluation agents metrics