Bryan Teo

Led end-to-end architecture, development and deployment of a voice booking system for salon services; built a real-time voice assistant with ElevenLabs Conversational AI, VAD, and queued TTS; designed modular chatbot orchestration with LangChain, OpenAI APIs, and Hugging Face models; implemented scalable RAG architecture with transformer-based embeddings, Pinecone, and structured context injection into LLM pipelines; built a hybrid Django-FastAPI backend; optimized APIs with WebSocket streaming, Redis caching, and Celery-based queues; established CI/CD, containerization with Docker, and deployment on Kubernetes across AWS/GCP; implemented privacy-first data pipelines with RBAC and encryption to support HIPAA and PCI DSS compliance.

Designed a real-time conversational AI video agent with multimodal models for natural human–AI interaction; built streaming speech-to-speech pipeline using Whisper-based ASR, GPT-3.5 dialogue models, and neural TTS; implemented an audio-to-visual animation pipeline with Wav2Lip and SadTalker to generate talking-head responses in real time; developed real-time visual perception pipelines (OpenCV, YOLOv5, MediaPipe); created embedding-based semantic retrieval (sentence-transformers, all-MiniLM-L6-v2) and PGVector for context-aware retrieval; optimized GPU inference and end-to-end latency; architected a low-latency backend using FastAPI, Redis Streams, and WebRTC with automated validation via Pytest.