Position Overview We are looking for an experienced Speech & Language AI Engineer to architect and deploy cutting-edge Speech, NLP, and Large Language Model (LLM) technologies for an AI-driven in-vehicle Virtual Assistant. This role focuses on delivering localized experiences for the Vietnamese, Indonesian, Arabic (Middle East), and Indian markets, ensuring high-performance conversational AI in real-time embedded automotive environments.
Key Responsibilities
- End-to-End Model Development: Research, develop, and fine-tune AI models (ASR, TTS, NLP, LLM) specifically optimized for Vietnamese, Indonesian, Arabic, and Hindi.
- Technical Optimization: Implement wake-word detection and enhance model inference for Cloud, Edge, and resource-constrained in-vehicle hardware.
- Production Deployment: Collaborate with R&D and Product teams to integrate models into scalable production systems.
- Lifecycle Management: Build and maintain automated data pipelines, rigorous evaluation frameworks, and proactive monitoring systems.
Requirements
- Experience: 35+ years of hands-on experience in AI/ML (Speech, NLP, or LLM), with a proven track record of deploying models for the Indian (Hindi) or Arabic markets.
- Domain Expertise: Deep understanding of the full-stack voice assistant pipeline: Wake-word ASR NLU/LLM TTS Command Execution.
- Deployment Mastery: Experience shipping production-grade AI systems in embedded, IoT, or automotive environments.
- Operational Excellence: Ability to own the entire model lifecycle (Train Eval Deploy Monitor Retrain).
Technical Skills
- Core Tech: Expertise in Python and deep learning frameworks (PyTorch/TensorFlow).
- State-of-the-Art Frameworks: Proficient with Transformers, HuggingFace, vLLM, LangChain, Whisper, Conformer, Zipformer, and wav2vec.
- Efficiency Techniques: Advanced knowledge of model compression (Quantization, Distillation, LoRA, Pruning) and latency optimization.
- LLM Innovations: Strong experience in Prompt Engineering, Chain-of-Thought, and RAG (Retrieval-Augmented Generation).
- MLOps: Familiarity with CI/CD for ML (MLflow, Weights & Biases) and telemetry-driven improvements.
- Linguistic Engineering:
- Handling dialectal variations (e.g., North vs. East Hindi, Arabic regional dialects).
- Experience with Code-switching (e.g., Hindi-English mix) and Slang/Conversational nuances.
- Managing morphological complexity and diacritics in Non-Latin scripts.
Language Requirements
- Native or Bilingual proficiency in Hindi, Vietnamese, Indonesian, or Arabic.
- Professional working proficiency in English (technical documentation and global collaboration).
Work Location & Contact
- Office: Technopark Building, Da Ton, Gia Lam, Ha Noi, Vietnam.
- Contact: Zalo 0769288088 | Email: [Confidential Information]
