Backend Engineering
Why RAG Systems Feel Dumb: The Hidden Cost of Poor API Boundaries
RAG systems often feel slow, inconsistent, or unreliable. This post explains how poor API boundaries around model calls create these issues and how to fix them.
Blog
Thoughts on AI engineering, backend, frontend, and building modern software.
AI Engineering
From Prototype to Production: Structuring AI Outputs as API Contracts
How to turn experimental prompts into reliable AI services with structured outputs and fallback strategies
AI Agents
Embeddings as Database Indexes: Tool-Call Permissions in AI Agents
How tool-call permissions and approval boundaries shape reliable AI agent systems using embeddings as the new database index
Software Architecture
Designing Modular AI Systems for Scalable Production
How to architect AI systems with modular design, internal APIs, and clear boundaries for reliable, maintainable production deployments
Security
Scoped API Keys for AI: Preventing Unauthorized Access in Production
Scoped API keys limit access to specific AI features, reducing risk from unauthorized usage and misconfigurations in production systems.
Prompt Engineering
Prompt Contracts as API Contracts: Structuring AI Outputs for Production Reliability
How structured prompt contracts prevent ambiguity in AI systems, ensuring reliable and predictable outputs in production environments
MLOps
The Hidden Cost of Quiet Prompt Changes: How Small Adjustments Trigger Big LLM Spend
A senior engineer's guide to tracking subtle prompt tweaks that inflate cloud costs and break production systems without raising alarms
Data Engineering
Data Freshness: Avoiding Stale Inputs in AI Pipelines
Data freshness is critical for AI systems. Learn how to avoid stale inputs and prevent hallucinations in production pipelines.
Cloud & Infrastructure
Deployment Boundaries for AI Services: Avoiding Hidden Infrastructure Costs
AI applications often hide infrastructure costs in deployment boundaries. Learn how to manage GPU, storage, and networking costs with practical strategies for production systems.
Backend Engineering
API Boundaries for AI: Designing Reliable Model Call Endpoints
How to design robust backend APIs for AI models with clear boundaries, timeouts, and error handling.