Writing
Notes from the lab
Challenges, decisions, and diagrams — RAG, agents, concurrency, models.
- Spring Boot · ProfilingSpring Boot Profiling: Finding the Real Bottleneck
p95 climbed and pods got fat. How we separated CPU hotspots from memory pressure — async profiles, heap clues, and what we changed after.
2026-03-20 · 10 min
- RAG · AgentsRAG That Earns Citations: What Broke Before Agents Got Useful
We said we built RAG. Then support answers drifted. Here is the chunking, hybrid retrieval, and groundedness loop that made citations trustworthy.
2026-03-12 · 11 min
- MongoDB · BackendThe MongoDB Race We Found Without Transactions
Two workers, one document, lost updates. How we detected a classic race on a payroll-adjacent status field — and fixed it with conditional writes.
2026-02-18 · 9 min
- Local LLM · Frontier modelsLocal LLMs vs Frontier Models: Where Each One Wins
When I run models on my own box, when I call Claude or GPT, and how the eval harness decides — not the hype cycle.
2026-01-28 · 8 min
- Agents · EvaluationAgent Evals Before Demo Day: What We Measure Now
Fluency lied to us. Process metrics, citation checks, and failure transcripts — lessons aligned with how serious agent teams grade systems.
2025-12-04 · 10 min
- Kubernetes · AWSKubernetes Notes from a Payroll Platform on EKS
OOM cascades, CoreDNS friction, and a backward-incompatible rollout — what we changed after each incident.
2025-11-02 · 7 min