Radhika User

I build AI-powered backend systems for real-time experiences, with 3+ years of experience creating LLM-powered applications, voice agents, and scalable APIs. I’m especially effective at turning modern LLM capabilities into production-grade solutions with reliable orchestration.

In recent work, I designed and developed voice agents using VAPI and Twilio, integrating LLMs for real-time speech-to-text, response generation, and voice synthesis. I built FastAPI-based backend services for conversation workflows, applied prompt engineering and context management, and routed LLM traffic through an LLM Gateway layer for provider fallback, caching, and cost/token tracking.

I also standardize agent-to-tool integrations by adopting Model Context Protocol (MCP), enabling cleaner access to tools and resources across systems. I’ve implemented retry logic, scheduling, and low-latency conversation pipelines to improve call success rates and overall reliability.

Beyond voice, I integrate LLMs and APIs (e.g., OpenAI GPT models) into internal tooling, and I develop secure, high-performance web applications with Python and Django. My focus stays on scalable system design, safe/validated outputs (e.g., RAG evaluation and Guardrails), and dependable cloud deployments on AWS.