Job Opportunities API

At Toku, we create bespoke cloud communications and customer engagement solutions to reimagine customer experiences for enterprises. We provide an end-to-end approach to help businesses overcome the complexity of digital transformation and deliver mission-critical CX through cloud communication solutions. Toku combines local strategic consulting expertise, bespoke technology, regional in-country infrastructure, connectivity, and global reach to serve the diverse needs of enterprises operating at scale. Headquartered in Singapore, Toku supports customers across APAC and beyond, with a growing footprint across global markets.   As a Founding AI Engineer, you will lead the development of our speech recognition capabilities, including contributing to open-source models optimised for APAC languages and telephony environments. You will own the entire machine learning pipeline from model architecture through to deployment, and publication on Hugging Face and GitHub. This is a unique opportunity to build technology that will serve billions of people across the Asia-Pacific region and beyond. What you will be doing   Model Development & Training Design and implement telephony-optimised speech recognition models for APAC languages (English variants, Mandarin, Thai, Vietnamese, Indonesian, and more) Develop comprehensive AI model training frameworks using PyTorch on local and cloud GPU infrastructure Create and optimise data augmentation pipelines addressing telephony-specific challenges (8kHz audio, codec artefacts, background noise, SNR optimisation) Build models that handle code-switching common in APAC contexts (Singlish, Hinglish, Taglish) APAC-Specific Optimisation Address tonal language challenges for Mandarin, Thai, Vietnamese, and other tonal languages Optimise for regional accent variations across target markets Develop evaluation benchmarks specific to APAC telephony contexts, including SNR and audio quality metrics Implement techniques for low-resource language support Infrastructure & Deployment Build scalable inference systems for real-time and batch processing Create containerised applications for model demonstration and testing Develop APIs for integration with telephony systems Deploy models on local and cloud GPU infrastructure Integrate with Toku's existing Llama 8B deployment for language model capabilities Open-Source Contribution (Future) Contribute to the preparation of open-source releases Write comprehensive technical documentation and user guides Conduct performance benchmarking and validation studies Contribute to the broader speech recognition community through publications and presentations   We’d love to hear from you if you have   Required Qualifications Bachelor's or Master's degree in Computer Science, Engineering, or related technical field with strong ML foundations 1-3 years of hands-on experience in machine learning projects Excellent Python programming skills Experience with PyTorch and deep learning model training Proficiency in handling large datasets and data preprocessing Understanding of speech processing concepts and techniques Experience with cloud platforms and GPU computing Familiarity with containerisation (Docker) and deployment practices Preferred Qualifications Portfolio of AI projects (open-source contributions highly valued) Familiarity with OpenAI Whisper and transformer-based architectures Previous experience with speech-to-text or audio processing projects Experience with open-source project development and collaboration Strong technical writing and documentation skills Familiarity with at least one APAC language's phonological characteristics Understanding of telephony audio characteristics (8kHz sampling, codec artefacts, SNR considerations) Publication history in speech recognition or related fields Personal Attributes Independen