Writing

Notes from production.

Deep dives on RAG isolation, FinOps, and backend work that makes SaaS scale. Written from real engagements, not tutorials.

Featured

▶

AI / RAG·Jul 2026

Eliminating Cross-User Data Leakage in Multi-Tenant RAG Pipelines

Why most multi-tenant RAG implementations leak data, and how enforcing per-user isolation at the SQL query layer — before vector similarity searches — closes the gap completely. Deep dive into pgvector + PostgreSQL RLS with code samples from a production FastAPI backend.

12 min readRead article

FinOpsJun 2026

72% Cloud Cost Reduction: A FinOps Playbook for SaaS

How I cut a client's AWS bill by $30K/month without touching feature velocity. Multi-region workload shifting, spot-instance scheduling, and the CDN tuning that nobody talks about.

9 minRead

AI / RAGJun 2026

NVIDIA NIM in Production: Latency, Cost, and Hallucination Trade-offs

Benchmarks from running nv-embed-v1 and llama-3.1-70b-instruct behind a FastAPI token-streaming layer. What worked, what didn't, and the gotchas that only show up at scale.

11 minRead

BackendMay 2026

Refactoring a Monolith into Microservices Without a Big-Bang Rewrite

The phased migration playbook I used to cut P99 latency by 33% while keeping the legacy API live. Feature-flagged parallel runs, contract testing, and the rollback gateway that saved us twice.

8 minRead

BackendMay 2026

PostgreSQL RLS + AsyncLocalStorage: Defense-in-Depth for SaaS

Tenant context belongs in more than one place. Here's the pattern I use to propagate it through NestJS request scope and into PostgreSQL Row-Level Security policies — so even a missed WHERE clause can't leak data.

10 minRead

FrontendApr 2026

Next.js 16 at 10K DAU: The Dashboard Optimization That Cut Loads by 33%

Composite indexes, N+1 batching, and React.memo against the right boundary. A practical story about identifying the one table that was killing dashboard load times.

7 minRead