Description
At Microsoft AI, we are at the forefront of technological innovation, creating powerful tools and products that transform the way people live and work. We are seeking passionate and talented engineers to join our multimodal team. As part of our mission, you will play a key role in training, evaluating and deploying state-of-the-art models shaping the future like MAI-Voice and MAI-Transcribe . Our team is small, fast-paced, and collaborative, committed to excellence in every aspect of our work. We value creativity, efficiency, and a user-first approach, ensuring that every decision we make is grounded in real-world evidence. This is an opportunity to work on products with immediate and far-reaching impact, placing you at the core of a rapidly evolving AI landscape.
Responsibilities: Model Training & Evaluation: Design and maintain training data “recipes” (data sourcing, cleaning, labeling workflows, QA, versioning, and lineage) and develop evaluation frameworks (gold sets, challenge sets, human-in-the-loop evals, regression suites). Drive continuous quality improvements through systematic error analysis, ablations, and experimentation to improve model performance, robustness, safety, and reliability. Training & Inference Optimization and Scaling: Optimize end-to-end training and inference performance to meet latency, throughput, cost, and reliability targets. Profile bottlenecks (data loading, preprocessing, GPU utilization, kernel efficiency), implement optimizations (batching, quantization, mixed precision, caching, model distillation, efficient serving patterns). Collaboration: Work closely with with other members of the AI research team, including researchers, engineers and product managers to define requirements, scope projects, and deliver high-impact solutions. Culture & Values: Actively contribute to a positive, inclusive, and collaborative team culture, embodying our Culture and Values .
Qualifications Required Qualifications Master’s degree in computer science OR equivalent technical experience. PhD is a plus Preferred Experience Experience building and maintaining training data pipelines (ingestion, cleaning, labeling, QA, versioning, lineage) Experience designing eval frameworks and datasets (gold sets, challenge/adversarial sets, human-in-the-loop evals, A/B tests, regression suites) and running error analysis. Strong experimentation skills: ablations, hypothesis-driven iteration, tracking metrics and qualitative rubrics. Dedication to writing clean, maintainable, and well-documented code with a focus on application quality, performance, and security. Demonstrated interpersonal skills and ability to work closely with cross-functional teams, including product managers, designers, and other engineers. Passion for learning new technologies and staying up to date with industry trends, best practices, and emerging technologies in web development and AI. Ability to work in a fast-paced environment, manage multiple priorities, and adapt to changing requirements and deadlines. Contributions or interest in audio