Akash Kumar

I’m an Applied AI and backend engineer focused on building production-grade GenAI systems for voice, chat, and enterprise automation. I specialize in RAG/GraphRAG pipelines, semantic retrieval, and LLM orchestration to deliver reliable, high-volume AI experiences.

In my current role, I’ve built and deployed GenAI workflows across voice and chat channels, automating 16+ enterprise use cases and reducing manual operations by 85%. I also designed low-latency real-time voice AI infrastructure using streaming ASR, LLM, and TTS over WebRTC and WebSockets—cutting response latency by 60% to under 700ms across 50,000+ monthly calls.

I’ve paired AI systems with practical backend design by building configurable eligibility and bundle-resolution logic from KYC, banking, employment, and document intelligence signals. This increased low-risk loan approvals from 10% to 25% while reducing false escalations—turning models into measurable business outcomes.

Earlier, I delivered FastAPI REST APIs for product, order, and payment domains, optimizing PostgreSQL logic and containerized deployments to reduce latency by 30%. I also strengthened my engineering fundamentals through Java/Spring Boot and Hibernate REST development, improving backend capacity by 30% and reducing response time by 25%, and I keep this craft reflected in projects like an enterprise voice calling agent and an open-set document identification system.