Job Description

Python ML Engineer (TTS / ONNX / Real-Time Audio)

Build voice AI that actually ships.

What you’ll be doing

You’ll work with our team to develop our Text-to-Speech (TTS) stack end-to-end.

• Train & fine-tune neural TTS models (Kokoro or similar)
• Make voices sound human — prosody, pronunciation, flow
• Turn research models into production systems
• Convert models to ONNX and squeeze every ms out of inference
• Optimize for real-time latency (not just accuracy on a notebook)
• Work on audio pipelines: text normalization → phonemes → spectrogram → waveform
• Ship APIs/services that run reliably under load

This is not a “train a model and call it a day” role.
You’ll be expected to make it fast, stable, and production-ready.

Job Requirements

Who we’re looking for

We care more about ability than years.

• You write solid Python
• You’ve used PyTorch (projects, internships, personal builds all count)
• You’ve trained something (NLP, CV, audio—doesn’t have to be TTS yet)
• You like figuring things out from papers / repos and making them work
• You care about performance (speed, memory, efficiency)

You might be:

• A junior engineer who actually builds things
• A self-taught ML hacker with real projects
• Someone who’s tired of theory and wants to ship real systems

Bonus points if you’ve touched

• TTS / speech models / audio processing
• ONNX / TensorRT / model optimization
• Quantization, pruning, or inference tuning
• Real-time systems or streaming pipelines
• Multilingual data (especially ASEAN languages)

What you’ll learn (fast)

• How to turn ML models into low-latency production systems
• How to optimize models beyond “it works”
• How real-world voice systems are built end-to-end
• How to balance quality vs speed vs cost

Environment

• Fast-moving, no bureaucracy
• High ownership (you build it, you own it)
• No hand-holding—but lots of real learning
• Focused on shipping, not slides

How to stand out

Skip the generic CV. Show us:

• GitHub projects
• Anything you’ve trained / deployed
• Audio / ML experiments
• Systems you’ve optimized

If you can prove you can build, we don’t care if you’re junior.

Skills