• Design, build, evaluate, and improve machine learning training and inference pipelines that power new AI-driven music experiences and help take them to fully scaled production-ready features. • Apply machine learning and prompt engineering knowledge across complex ML pipelines to support rich user experiences involving large language models. • Create evaluation frameworks, including LLM-as-judge pipelines, to measure quality and build fast feedback loops that enable rapid and confident iteration. • Partner with music subject-matter experts to bootstrap training and reference data, including synthetic generation, expert curation, and taxonomy design. • Build scalable systems that balance experimentation velocity with production rigor, ensuring strong performance, reliability, and latency at Spotify scale. • Collaborate closely with Data Science teams to connect evaluation frameworks with real-world usage signals and continuously improve model quality. • Contribute to technical direction and engineering best practices across model deployment, observability, experimentation, and production infrastructure. • Work cross-functionally with engineering, product, design, and music industry partners to shape entirely new listening experiences for artists and fans.

• Design, train, fine-tune, and evaluate deep learning, classical ML, and foundational GenAI models to drive core product features. • Build, scale, and maintain robust ML pipelines for continuous training, evaluation, and real-time/batch inference using modern MLOps frameworks. • Collaborate with backend and frontend teams to integrate AI models into microservices, ensuring low latency, high availability, and optimal resource utilization. • Architect scalable data pipelines for preprocessing, vectorizing, and ingestion of massive structured and unstructured datasets. • Implement rigorous evaluation frameworks for model alignment, bias mitigation, guardrailing, and cost/latency optimization (e.g., quantization, distillation). • Provide technical leadership, conduct thorough code reviews, and mentor junior/mid-level engineers on best practices in software craftsmanship and ML engineering.

• Build, train, and optimize computer vision models for image classification, face liveness detection, and presentation attack detection (PAD) / anti-spoofing • Work on real-world identity verification and biometric authentication problems, improving model performance on noisy, adversarial inputs such as spoofed images, replay attacks, deepfakes, and synthetic media • Design and run experiments to improve model accuracy, recall, robustness, and fraud detection performance using techniques such as augmentation, class balancing, architecture tuning, and hard-negative mining • Design, train, and improve deep learning models (e.g., CNNs, Vision Transformers, and foundation models), including loss function design, hyperparameter optimization, and performance tuning on large-scale image datasets • Prepare and curate large, noisy datasets, including data ingestion, validation, cleaning, deduplication, labeling strategies, and dataset QA to improve model reliability and generalization • Develop evaluation protocols and success metrics that balance fraud detection effectiveness, false acceptance rates, false rejection rates, and overall business impact • Develop production-grade training and inference pipelines on AWS with strong reproducibility, monitoring, observability, and cost controls • Productionize models as resilient Python services and libraries; collaborate with platform teams to optimize APIs, latency, scalability, and operational reliability • Contribute to the evolution of our Identity Verification (IDV) platform by modernizing legacy components and improving model performance, maintainability, and modularity • Partner closely with Product, Customer Success, Fraud, and Platform Engineering teams to ensure ML solutions meet privacy, compliance, security, and reliability requirements • Support and mentor other engineers through design reviews, code reviews, experimentation best practices, and knowledge sharing • Research and evaluate emerging techniques in face liveness detection, presentation attack detection (PAD), deepfake detection, biometric authentication, and adversarial machine learning to strengthen our fraud prevention capabilities

• Tune and optimize ML/DL models for production-scale audio deepfake detection. • Investigate failure cases in the client environment, build custom evaluation frameworks, and implement mitigation strategies for model robustness. • Drive model iteration for performance under a variety of real-world environments, e.g. compression artifacts, noise, telephony, and streaming pipelines. • Present technical findings and model performance insights to internal stakeholders. • Interface with Product and Engineering teams to build a deep understanding of the production environment and incorporate relevant evaluations for performance assessment.