Embodied AI & Multimodal LLM Engineer.

Job Description

As an Embodied AI and Multimodal LLM Algorithm Engineer, you will build ChipTalk's next-generation full-duplex streaming voice brain (enabling sub-800ms natural voice interruption), multimodal interaction agents, and edge-cloud collaborative decision engines, empowering companion toys, AR glasses, and smart wearables with unprecedented emotional connection and situational awareness.

Key Responsibilities

• Develop and optimize full-duplex streaming conversational pipelines integrating on-device VAD, streaming ASR, emotional TTS, and low-latency LLMs
• Architect embodied agent frameworks featuring episodic long-term memory, emotional state machines, and dynamic tool-calling integrations
• Design hierarchical edge-cloud collaborative intelligence: on-device SLMs for microsecond intent classification and emergency safety; cloud LLMs for complex reasoning
• Implement multimodal perception combining first-person video (POV), IMU telemetry, and proximity sensing for contextual awareness
• Track, evaluate, and fine-tune leading open-source models (LLaMA, Qwen, MiniCPM, Whisper) for vertical hardware deployments

Requirements

• Master's or Ph.D. degree in Computer Science, AI, Robotics, or Computational Linguistics
• 3+ years of engineering experience in LLMs, conversational speech AI, or embodied intelligence
• Deep proficiency in PyTorch, Transformer architectures, attention optimizations, and end-to-end streaming speech systems
• Solid understanding of autonomous agent patterns (ReAct, state graphs) with proven hands-on fine-tuning (LoRA, SFT) experience
• High sensitivity to end-to-end latency; familiar with WebSocket/WebRTC streaming protocols and concurrent systems
• Passionate about embodied AI, companion robots, and next-generation human-machine intimacy