To apply for this job, you need to complete both steps below:
STEP 1:
Please click the link to submit your application directly to the company:
Your application will only be received by Recruiter if submitted via above link.
STEP 2:
Kindly scroll to the bottom of this page and complete the short VinUni Tracking Form.
Filling out this form alone does not count as applying. Kindly remind this form is not part of the company's application process. It only helps Careers, Alumni, Industry and Development (CAID) Department discover more opportunities and follow up in case of system issues.
Responsibilities
- Research and build a unified, real-time S peech U nderstanding model - powered by an LLM - that lets the Voice Agent perceive speakers, their state, the environment, and audio content beyond the transcript; prototype agent-awareness behaviors using simulated agents and scripted scenarios.
- Build the Vietnamese data and benchmark foundations: data collection, annotation, and synthetic data, plus a shared evaluation framework that measures real impact on agent behavior, not only model metrics.
- Train, fine-tune, and post-train audio and audio-LLM models (pretraining, SFT, alignment ) and optimize them for low-latency streaming inference.
- Work with the ASR, TTS, Foundation, and Agent teams to integrate models into conversational systems and voice assistants, including demanding in-vehicle conditions.
- Stay current with the research community and contribute new ideas and, where appropriate , publications.
In this role, you'll join the Speech Understanding team and work across the full loop - data, models, evaluation, and integration into a production Voice Agent. We're looking for people with relevant experience, passion, and drive, ready to learn fast and thrive under pressure.
Requirements
- Bachelor's Degree or higher in Computer Science, Artificial Intelligence, or related fields.
- Minimum 2 years of experience in AI/ML, with a focus on speech and audio (ASR, audio understanding, or related areas) or on audio/multimodal LLMs.
- Proficient in Python and PyTorch, with a strong grasp of Transformers and modern speech/audio models (self-supervised encoders, Whisper-style models, audio-LLMs).
- Hands-on experience training or fine-tuning models - ideally LLMs (SFT, PEFT/LoRA) - and building or evaluating speech datasets.
- Comfortable with open-ended, research-driven problems where the problem definition is still evolving.
- Good English proficiency to read technical documentation and research papers.
Preferred
- Experience with Vietnamese, multilingual, or code-switching speech.
- Experience with real-time streaming systems or voice agents, and interest in multimodal (voice and vision) agents.
- Publications in top-tier AI/speech conferences (e.g., Interspeech, ICASSP, NeurIPS).