suchandra siripuram

I’m an AssociateAI Engineer building predictive matching and AI safety infrastructure at QuilrAI. I designed a priority-based cascade system that combines EDM, Falcon semantic embeddings, and few-shot intent classification, hitting 91.3% accuracy with <5ms latency using pre-computed embeddings and optimized SQLite/LanceDB storage.

I also built a custom FastDeploy model serving layer with consumer-producer batching to support high-throughput concurrent inference. In production, I deploy ONNX-accelerated safety models like Prompt Guard and Vectara hallucination detection for AI systems.

On the data side, I engineered a cross-cloud extraction pipeline across AWS, GCP, and Azure with unified schemas and automated transformation logic. That work centralized governance and security auditing for 100+ AI resource types spanning services like Bedrock and SageMaker and Azure RAG.

Before that, I interned at Wavelabs Technologies, where I engineered an automated data processing pipeline for Kubernetes deployments using JavaScript, Node.js, and PostgreSQL. I optimized deployment cycle time from 6 weeks to 3 weeks and partnered on real-time monitoring and data validation.