Convin Hybrid SDE 1 (Full stack) at Convin

Built Convin’s Voice AI platform integrating telephony providers (Ubona, Agami, Exotel), STT (Deepgram), TTS (Cartesia, ElevenLabs), and multiple LLMs. Reduced audio latency by ~30% using interruption handling and audio chunking. Designed and shipped scalable FastAPI backend APIs for AI-driven campaign orchestration (agent creation, campaign lifecycle, structured payload handling, multi-tenant isolation). Implemented distributed async pipelines using Celery, Redis, and RabbitMQ for scheduling, call execution, and webhook reliability to support 100,000+ calls at production scale. Improved LLM agent reliability with voicemail detection, turn-taking support, prompt customization, and multilingual tool-parameter handling; integrated streaming TTS over SSE with low-latency chunk processing and synthesis observability. Owned end-to-end frontend across a large TypeScript/React codebase (230+ commits) covering campaign flows, analytics dashboards, and shared config via Redux/RTK and typed do