I architected and shipped a real-time AI voice agent platform using Python, FastAPI, Django, LiveKit, WebRTC, and Twilio on GCP. The system was designed for scalable, low-latency inference and bidirectional audio streaming.

I integrated Gemini 2.5 Pro, Vertex AI embeddings, Whisper, Deepgram STT, and ElevenLabs TTS into multimodal RAG and streaming inference pipelines. I also built interrupt-driven agent logic, turn-taking, barge-in detection, and context-aware state management for natural conversational flow.

I developed audio ingestion, preprocessing, batch inference, and evaluation workflows, and collaborated with clinicians, researchers, and product teams to improve accuracy and reliability. I also mentored junior engineers and helped drive cross-functional delivery of new AI capabilities.

I delivered full-stack AI applications using React, Next.js, TypeScript, Python, FastAPI, Django, Node.js, Redis, PostgreSQL, AWS, and GCP. My work supported internal AI platforms and large-scale user-facing retrieval systems.

I built and productionized multimodal, multilingual embedding-based retrieval models serving 20M+ users with low-latency and high-QPS requirements. I also designed cross-lingual contrastive learning pipelines, synthetic data generation workflows, and conversational search systems with query rewriting and RAG grounding.

I led observability and governance efforts using LangSmith, Langfuse, PostgreSQL, and custom telemetry dashboards, and I optimized FAISS indexes, Redis-backed serving, ONNX deployment, quantization, and Triton inference.

I led development of a clinical decision support platform that enabled natural language search across treatment guidance and patient insights. The work improved clinicians' ability to find relevant information quickly.

I built a large-scale document retrieval system indexing over 1M clinical documents using hybrid search, semantic ranking, vector retrieval, and contextual grounding. I also developed workflow automation services for retrieval, summarization, and compliance-driven processing.

I created scalable backend services and clinician-facing applications using Python, FastAPI, Go, React, Angular, TypeScript, and Redux, improving performance, accessibility, and workflow efficiency.

I built and optimized platform services on AWS using Node.js, Nest.js, Terraform, Java, Spring Boot, and .NET Core. My work improved deployment reliability, reduced downtime, and increased API performance.

I also developed responsive frontend applications with React, Angular, and Next.js, created reusable UI component libraries, and integrated applications with backend services through REST, GraphQL, and Apollo.

In addition, I helped design media-processing pipelines with Delta Lake, AWS Lambda, and FFmpeg, and I optimized Node.js services and AWS compute resources to reduce costs and improve rendering speed.