ML Scientist Intern

Job Description Our team is responsible for helping Customer Experience teams to achieve their best, by intelligently solving repetitive work, so they can shift their focus to solving more sophisticated problems. We use the latest trends in Machine Learning and AI algorithms to help us on that mission, and we're passionate about empowering our customers. As a Machine Learning Scientist Intern, you will drive development, evaluation, and deployment of novel ML/AI models to power intelligent automation and customer service solutions at scale. You will collaborate closely with engineers, product managers, and cross-functional teams to translate research into solutions directly impacting millions of support interactions. What you get to do every day Research, prototype, and develop state-of-the-art NLP/ML models for use cases such as intent detection, auto-assist, chatbots and intelligent agent routing. Design and execute rigorous experiments and evaluations (offline/online, A/B) to improve model accuracy and robustness. Work closely with ML Engineers to productionize ML solutions—including data pipelines, scalable model serving, and monitoring. Analyze large, multi-lingual customer interaction datasets to uncover insights and power new solutions. Participate in technical reviews and share knowledge of underlying ML methodologies and best practices. Present your work to a multi-disciplinary, global audience. Stay up to date with recent literature in Machine Learning and Natural Language Processing (NLP) and share knowledge internally. Key challenges / use cases How do we enrich customer service conversations with accurate language detection, intent recognition, and real-time sentiment analysis, to enable proactive customer engagement and optimal routing? How can we automate all customer service interactions as much as possible, from process automation to agent assistance and chatbots with a knowledge base? How do we optimize routing at scale—matching tickets or chats to the most appropriate agent/team in real-time across multiple languages and regions? How do we automate large-scale A/B testing and model evaluation (online and offline) to continually iterate and improve ML-driven triage and agent-assist tools? What novel approaches or architectures (e.g., retrieval-augmented generation, few-shot/fine-tuning strategies) can extend our conversational AI platforms to unlock new customer support use cases and modalities? How do we efficiently operationalize, monitor, and update large-scale (LLM/ML) models in dynamic, high-throughput production settings, ensuring model health, drift detection, and continuous learning? How do we combine signals from conversation context, customer history, and external data to improve prediction and decision accuracy across our ML services? What are the emerging advancements in ML/AI research (e.g., large language models, efficient adaptation, re-ranking, retrieval, or explainable AI) that should be incorporated into Zendesk’s customer experience ecosystem? How can we bridge the gap between cutting-edge research and impactful product features, rapidly validating ideas in production and quantifying their real-world business value? And many more! What you bring to the role MSc (or PhD) degree in computer science, electrical engineering, math, or related areas. A good foundation in statistics and machine learning techniques. Solid coding skills in Python; experience with ML frameworks (preferably PyTorch). Experience with deep learning and/or NLP is a bonus. Great written and verbal communication skills. A collaborative and can-do attitude. A desire to learn and to grow. What our tech stack looks like Our code is written in Python and Ruby. Our servers live in AWS. Our machine learning models rely on PyTorch. Our ML pipelines use AWS Batch and MetaFlow. Our data is stored in S3, RDS MySQL, Redis, ElasticSearch, Snowflake and Aurora. Our services are deployed to Kubernetes ... (truncated, view full listing at source)