Writing
Hands-on essays on building AI systems, cloud platforms, and the distributed services underneath them. No hype — just what works and what breaks.
RAG, evaluation, and shipping LLM features to production.
A RAG system without an evaluation harness is a system you can't safely change. A field guide to retrieval vs. generation metrics, public benchmarks like BEIR and RAGBench, and frameworks (RAGAS, TruLens, DeepEval) for turning evals into a CI gate.
6 min read
Why most RAG demos break in production, and the retrieval, evaluation, and guardrail patterns that make retrieval-augmented generation dependable at scale.
1 min read
Cloud-native foundations, Kubernetes, and platform design.
A component-by-component walk through Oracle Cloud's Virtual Cloud Network — with the AWS VPC mappings called out where the divergence actually matters.
10 min read
A pragmatic approach to building an internal developer platform on Kubernetes without collapsing under the weight of its own abstractions.
1 min read
Distributed systems, consistency, and scaling trade-offs.
A comprehensive guide to building a production-grade AI e-commerce platform on Oracle Cloud. The architecture's heart is a shared data spine — a product graph and customer 360 kept fresh by change-data-capture — that every AI capability reads from. From discovery (search, recommendations) through conversion (shopping assistant, personalization) to operations (demand forecasting, dynamic pricing, fraud), every capability plugs into the same spine, with the evaluation, logging, and cost model that separate a demo from a production system.
11 min read
A system-design walk-through of a CDN, built layer by layer. Start from a single origin, discover why it collapses under load, and add only the mechanism each bottleneck demands — caching, the shield, cache keys, eviction, invalidation, and the tail.
8 min read
A system-design walk-through of an API gateway, built layer by layer. Start from clients calling services directly, discover why that breaks, and add only the mechanism each bottleneck demands — a single entry point, then auth, routing, rate limiting, and the discipline that keeps the gateway from becoming the bottleneck it set out to prevent.
9 min read
A comprehensive guide to building production-grade image search on Oracle Cloud, walked through three scenarios of growing scale. Start serverless with OCI Generative AI's managed embedding and vector store, grow into Autonomous Database for relational needs, and scale out with OpenSearch and self-hosted models — plus the evaluation harness, logging, and cost model that separate a demo from a production system.
11 min read
Event-driven architecture is powerful, but the trade-offs only surface under load. A field guide to consistency, idempotency, and operational sanity.
2 min read