Our NLP-based search, written in TypeScript with NestJS, currently needs about 30 seconds to answer a query—far too slow for our users. I need you to dig into the existing codebase, profile every stage of the pipeline, pinpoint the real bottleneck—whether it sits in model inference, database access, network latency, or something else—and tune or redesign it so typical responses land in five seconds or less. You will get full technical details and access to the relevant repos, benchmarks, and infrastructure once we start. I expect clear, side-by-side before/after metrics, well-documented changes, and any scripts or configs used for profiling or load testing. If you have proven experience lifting performance on NLP systems, especially in TypeScript or NestJS environments, I’d like to see it. A concise note about similar work you’ve done is all that’s needed; no elaborate proposals. Price is negotiable, but please apply only if you’re confident you can deliver the sub-5-second goal.