Speech AI Research Engineer

Our AI voice agents have already handled 40,000+ real patient calls for health tourism clinics. Now we’re bringing the speech stack in-house: our own recognition, our own voices, on our own cloud infrastructure. We need someone who wants to build that, not just read about it.

Fully remote. Part-time or full-time. University students welcome.

About volina.

Volina builds AI infrastructure for the future of health tourism. Our multilingual voice agents help healthcare providers manage patient conversations, automate communication and deliver more personal experiences at scale. We have active customers and 40,000+ completed calls. The next step is our own cloud-based Speech AI: accurate speech recognition and natural, human-like speech synthesis, with a sharp focus on Turkish and real-time conversation.

What you’ll work on

Speech Recognition (STT)

• Research, benchmark, fine-tune and improve speech recognition models
• Make Turkish recognition hold up across accents, speaking styles and messy real-world telephone audio
• Build and evaluate conversational speech datasets

Speech Synthesis (TTS)

• Explore and fine-tune neural TTS models for natural, expressive Turkish voices
• Push pronunciation, prosody, emotional expressiveness and real-time generation

Cloud Infrastructure & Deployment

• Help architect GPU-powered, cloud-native speech services and streaming inference pipelines
• Measure and improve latency, reliability, scalability and inference cost

Research & Experimentation

• Track new Speech AI research, evaluate open-source models and reproduce what looks promising
• Design measurable experiments and ship the winners to production with our engineering team

What we’re looking for

• Strong Python skills and solid machine learning and deep learning fundamentals
• Experience with, or strong interest in, PyTorch, Hugging Face and model fine-tuning workflows
• You can research a problem on your own, run experiments and write down what you found
• Genuine curiosity about speech technologies, generative AI and model architectures
• A collaborative mindset and the will to take ownership of hard problems

Nice to have

• Hands-on experience with Whisper, wav2vec 2.0, NVIDIA NeMo or neural TTS models
• Audio preprocessing, speech datasets and quality evaluation
• Linux, Docker, cloud platforms, GPU computing or streaming audio systems
• Personal ML projects, open-source contributions or research experience

What we offer

• Real research resources: dedicated GPU compute and a cloud budget for training, fine-tuning and experiments
• Technical ownership: help shape Volina’s speech infrastructure and its architectural decisions
• Freedom to explore: test new approaches and turn promising experiments into measurable product wins
• Direct product impact: your work shows up in live conversations with real customers
• Equity: a stake in the company you help build
• Remote flexibility: fully remote, part-time or full-time
• Team events & offsites: regular events that bring the team together
• Room to grow: applied research, model development, cloud infrastructure and production AI, all in one role

Who should apply

We care more about how you think, what you’ve built and how curious you are than your university, degree or years of experience. University students, recent graduates, self-taught developers and experienced engineers are all welcome. You don’t need to tick every box. Strong foundations in one area and the drive to learn the rest beat a perfect checklist.

How to apply

1. Apply through this LinkedIn post
2. Share your work: CV, GitHub, Hugging Face profile, or a technical project you’re proud of
3. Tell us what you explored: an ML experiment, a model you fine-tuned, or a research problem you dug into. We especially want to hear this one.

If you’ve ever fine-tuned Whisper at 2am just to see if it would work, we should talk.