Kamran Khan

At Nextyn, I architect AI-powered discovery and agentic platform infrastructure, headlined by a dual-engine hybrid search (Qdrant vector search + Meilisearch) across 35K–40K+ experts that accelerated recruiting turnaround by over 50%.

I engineered high-throughput indexing pipelines with async workers and Redis queues, slashing indexing latency from 60 seconds to 2 seconds and cutting end-to-end search query times from 8 seconds down to 2–3 seconds. I also automated data ingestion for 14K+ expert profiles alongside streaming financial feeds, while standardizing Docker and CI/CD deployment pipelines across four production microservices.

Outside of my day job, I design and ship independent agentic systems and developer tooling under Cadence Labs:

• DossierAI (Micro-SaaS): Built an autonomous account intelligence engine using Next.js 15, FastAPI, and LangGraph state machines that asynchronously crawls commercial domains to extract growth trigger signals, executive dossiers, and multi-channel outbound suites (dossier-ai.kamrankhan.xyz).

• ProbeAI: Published an open-source NPM CLI profiling engine that benchmarks latency, token throughput, and routing across multiple LLM providers.

• Recall: Engineered an autonomous bookmarking and retrieval engine using Next.js, FastAPI, and vector embeddings to extract, summarize, and semantically index web content.

Previously at vConstruct, I architected an enterprise OpenAI-powered document retrieval assistant, end-to-end user telemetry pipelines on Azure Application Insights, and modular React/TypeScript frontend systems that boosted runtime performance by 30% and reduced production bundle size by 22%.