Annual base salary: CAD $90,000 to $100,000
About the role
ALEBEX AI builds Canadian voice AI infrastructure. We are hiring a Junior AI Engineer to help develop and maintain our real time speech engine. You will work across speech recognition, end of turn detection, and speech synthesis, the components that let a voice agent hear a caller, know when to respond, and speak naturally.
You will work with our engineering team to design experiments, train and evaluate models, and bring improvements into production. The goal is to make voice conversations more accurate, responsive, and reliable.
What you’ll do
- Develop, train, and evaluate models for speech recognition (STT), end of turn detection, and speech synthesis (TTS).
- Help run pretraining and post training experiments. Depending on the model and the problem, these may involve supervised fine tuning (SFT), direct preference optimization (DPO), group relative policy optimization (GRPO), or on policy self distillation (OPSD).
- Prepare and check speech and text datasets for training and evaluation.
- Build evaluation tools to measure recognition accuracy, voice quality, turn taking, and response latency. Use the results to identify errors and test improvements.
- Write, test, and maintain Python code for training pipelines and model services. Help improve inference speed and troubleshoot issues that appear in production.
- Review relevant speech and language model research and test ideas that can improve the product.
What you’ll bring
- Strong Python skills and experience with PyTorch or a comparable deep learning framework.
- A solid foundation in machine learning and neural network training, with a project where you trained or adapted a model and evaluated the results.
- Experience writing, testing, and debugging code, and the ability to explain your experimental choices and findings.
- A degree in computer science, engineering, statistics, or a related field, or equivalent practical experience. Relevant coursework, internships, research, and independent projects count.
Nice to have
- Experience with speech or audio, including speech recognition, speech synthesis, audio processing, speech tokenization, or streaming systems.
- Exposure to GPU or distributed training, model serving, or inference optimization.
- Familiarity with one or more of the training methods listed above.
- A research paper, open source contribution, or project that shows how you approach an applied machine learning problem.
When applying, include a short note about a relevant project, what you contributed, and how you measured the results.