Senior Applied Machine Learning Scientist
Sanas is revolutionizing the way we communicate with the world’s first real-time algorithm, designed to modulate accents, eliminate background noises, and magnify speech clarity. Pioneered by seasoned startup founders with a proven track record of creating and steering multiple unicorn companies, our groundbreaking GDP-shifting technology sets a gold standard. Our initial deployment is laser-focused on elevating the standards of customer experience centers. Testimonials from our partners reveal staggering double-digit improvements in mission-critical KPIs, coupled with boosts in CSAT and NPS. More than just a tool, our technology champions a bias-free workspace. This not only fosters a positive work environment but has also been instrumental in reducing employee attrition and curbing training expenditures.
Sanas is a 100+ strong team, established in 2020. In this short span, we’ve successfully secured over $50 million in funding. Our innovation have been supported by the industry’s leading investors, including Insight Partners, Google Ventures, General Catalyst, Quiet Capital, and other influential investors. Our reputation is further solidified by collaborations with numerous Fortune 100 companies. With Sanas, you’re not just adopting a product; you’re investing in the future of communication.
We are seeking a passionate and skilled Applied Machine Learning Scientist with expertise in the speech AI domain to enhance and optimize our production AI systems. In this role, you will take a data-centric approach to identify and address areas where existing models underperform. Your contributions will directly improve the accuracy, efficiency, and robustness of our speech-to-speech systems, ensuring they meet the needs of diverse users in real-world environments.
This position requires a strong foundation in ML techniques, an innovative mindset, and a deep commitment to continuous improvement of deployed systems.
Key Responsibilities
Analyze performance gaps in production models, especially in specific domains, accents, languages, or noisy conditions.Iterate on existing models by refining architecture, hyperparameters, and training pipelines to address identified issues.Apply techniques like active learning, transfer learning, and fine-tuning to improve model generalization.Identify data deficiencies impacting model performance (e.g., data imbalance, low-quality annotations).Curate and preprocess datasets to improve model robustness and accuracy in underperforming areas.Collaborate with Data Engineering, MLOps, and Annotation teams to generate high-quality, targeted datasets.Develop and maintain automated tools to monitor model performance across diverse use cases.Conduct error analysis to understand model failure modes and devise actionable solutions.Define, track, and report metrics (e.g., WER) that measure model system performance.Stay updated with advancements in ML, applying cutting-edge techniques like self-supervised learning or domain adaptation.Test novel approaches to improve model scalability, speed, and accuracy.Work closely with Product, Engineering, and QA teams to ensure seamless deployment and integration of model updates.Clearly communicate findings, progress, and technical concepts to stakeholders. Must have qualifications
Master’s or Ph.D. in Computer Science, Electrical Engineering, or a related field with a focus on Machine Learning or Speech Processing.5+ years of hands-on industry experience in developing and implementing the following systems: Speech-to-textText-to-speechVoice conversionSpeech restorationSpeech translationProficiency in Python and frameworks such as TensorFlow, PyTorch, or Kaldi.Solid understanding of speech AI techniques, including acoustic modeling, language modeling, and evaluation.Hands-on experience with data preprocessing, augmentation, and error analysis for speech data.Experience with cloud-based technologies such as AWS, GCP, or Azure and deep learning training with GPUs. Preferred experience:
Proven track record of contributing to a scalable, commercial machine learning model in production.Proven track record of achieving significant results as demonstrated by publications at leading workshops, journals or conferences such as ICASSP, INTERSPEECH, or similar.Knowledge of tools for audio processing (e.g., Praat, FFmpeg).Experience with model performance optimization for edge devices.Understanding of MLOps practices for deploying and monitoring models in production.
Apply for this job →