PinnedAishwarya B Raghavan·Sep 17Race Conditions in FastAPI: Why async def Doesn't Save YouThe customer clicked “Cancel Order” once, and it was sad enough to see him cancel an order. But when we checked our payments table, it…A response icon2A response icon2
PinnedAishwarya B Raghavan·Jul 29Cross-Tenant Data Leakage: Query Scoping Alone Isn’t EnoughMulti-tenant SaaS platforms run hundreds of companies’ data through shared infrastructure. A cross-tenant leak can end a customer…A response icon2A response icon2
PinnedAishwarya B Raghavan·Jul 8How FastAPI Actually Runs Your ML Model EndpointServing an AI model over HTTP looks like a short job: load the model, wrap it in a FastAPI route, return the output. Under real traffic…
PinnedAishwarya B Raghavan·Jun 15What your query plan has been trying to tell youA debugging guide for slow queries, ignored indexes, ORM lazy loading, and lock contention in PostgreSQL
PinnedAishwarya B Raghavan·May 6React Hooks Field Guide: Every Hook, Every Tradeoff, React 19 IncludedEvery front-end developer has a hook they reach for first, almost out of habit. useEffect for anything async, useState for anything that…
Aishwarya B Raghavan·Nov 2FastAPI Under Load: Diagnosing Where The Requests WaitMost FastAPI performance advice focuses on fixes: adding workers, increasing pool sizes, or switching to a faster serializer. These changes…A response icon2A response icon2
Aishwarya B Raghavan·Oct 3Kafka Stratagems: Partitions, Consumers, and the Order You Scale InOur order-processing service takes an order, charges the card, and writes the result to the database. Between the API and the processor is…
Aishwarya B Raghavan·Sep 24When “Global” Isn’t Global: The Hidden Failure Mode of In-Memory State in FastAPIA rate limiter can look correct in development and still fail in production.
Aishwarya B Raghavan·Aug 6How I Manage 20+ Infrastructure Services with Docker Compose ProfilesYou don’t notice your local infrastructure growing until docker compose up starts taking longer than the work you're trying to do.
Aishwarya B Raghavan·Jul 16Your LLM Isn't an Orchestrator: An Agent Architecture That ScalesLLMs are bad at distributed transactions. Here's an agent architecture that splits tool exposure, policy, and coordination cleanly