Saurabh Yadav

I built a finance investigation agent that examines alerts across 590K transactions, retrieves similar cases from graph memory, scores risk, and routes actions under bank policy. A calibration test found two scoring bugs and cut false positives from 42.5% to 2.5%.

At FarAlpha Technologies, I built Python microservices on AWS Lambda, API Gateway, and ECS, with Redis caching and SQS processing. I also built an asynchronous invoice-processing worker and owned observability with OpenTelemetry tracing in CloudWatch and Grafana.

At DatenStrom-3AG Solutions, I fine-tuned an LLM for chatbot services and built backend services exposed to client apps through JWT-secured REST APIs. At Avani Enterprises, I shipped full-stack Next.js and Node.js apps on PostgreSQL from requirements through deployment.

On a team project at Hacker House Goa 2026, I worked on a multilingual voice assistant with speech-to-text, hybrid retrieval, and spoken answers in eight Indian languages. Its input, context, and grounding checks block hallucinated answers, and its latency was benchmarked against a 200ms budget.