My Role
- #2 Software Engineer at CrowdListen (January – May 2026), one of two engineers building the product.
- Worked across the stack: the React/TypeScript frontend, Python/FastAPI services, background workers, deployment, and the evaluation loop that decided what shipped.
What I Built
- Job pipeline. An analysis takes 60–90 seconds, so it cannot live on the request path. Requests enqueue jobs on Redis-backed queues, horizontally scalable workers pull from 11 sources, and progress streams back to the browser over server-sent events.
- Deployment. Containerized the API and worker services with Docker and deployed them to AWS ECS/ECR through GitHub Actions CI/CD, with health checks, rolling deployments, and autoscaling on queue depth.
- Reliability. End-to-end traces plus pytest and Vitest integration and load tests aimed at the failure modes we saw in practice: timeouts, retries, duplicate jobs, and dropped SSE connections.
- Retrieval. An LLM-routed retrieval pipeline on PostgreSQL/pgvector with hybrid rank fusion, semantic deduplication, and engagement-weighted clustering.
- Agent interface. Restructured a flat 30-tool MCP surface into five core tools plus on-demand modules, so an agent only loads the tools a task needs.
Results
- Relevance errors down 47%, end-to-end task success up from 72% to 95%, and the share of high-signal themes up 40%.
What CrowdListen Does
CrowdListen gathers customer, competitor, and community signals from sources like Reddit, YouTube, TikTok, X, review sites, and support channels, then structures them into themes, evidence, and recommended actions that product teams and their AI agents can use.