Member of Technical Staff - AI Data Platform, Frontier Models

Help build the world's most advanced multimodal dataset at Microsoft AI. The AI Data Infra team at Microsoft AI is responsible for building data infrastructure to help MAI teams generate the biggest and best training dataset. Work involves data pipelines, Spark, Ray, Vector Databases, and all other aspects of data infra. This role is part of Microsoft AI's Superintelligence Team (MAIST). Candidates are expected to work from a designated Microsoft office at least four days a week if they live within 50 miles (U.S.) or 25 miles (non-U.S.) of that location, subject to local law.

Responsibilities

Build AI Data Infrastructure and Data Engines: develop and evolve large-scale AI data infrastructure for data collection, ingestion, cleaning, curation, generation, governance, metadata management, query, analytics and hybrid search; Develop Intelligent Multimodal Data Processing Systems: lead automated understanding and processing of text, images, video, documents, build labeling and taxonomy systems, and design and train models for classification, recognition, captioning, quality scoring and related tasks; Build AI-Native Data Pipelines: leverage LLMs, VLMs and Agents to automate data collection, filtering, deduplication, quality diagnosis, annotation, generation, orchestration and anomaly detection; Build AI-Native Data Storage and Table Layers: design open, AI-optimized storage using Lance, Iceberg, Paimon and Parquet; Discover and Build Rare, High-Value Datasets: develop differentiated datasets for challenging domains using real-world data collection, human annotation and synthetic data generation; Drive the Data-Model-Evaluation Iteration Loop: use evaluation feedback and failure analysis to identify data gaps and establish metrics for data contribution to model capability; Partner Closely with Model and Training Teams: collaborate on training-data construction, data feedback loops, targeted data mining and evaluation-driven iteration

Requirements

Master's Degree in Computer Science, Math, Software Engineering, Computer Engineering, or related field AND 3+ years experience in business analytics, data science, software development, data modeling, or data engineering OR Bachelor's Degree in Computer Science, Math, Software Engineering, Computer Engineering, or related field AND 4+ years experience in business analytics, data science, software development, data modeling, or data engineering OR equivalent experience; Experience with distributed data Platforms such as Spark, Flink or Ray; Proficient experience in Python and experience with SQL and Shell; Experience with Multimodal Data (Text, Image, Video, or Audio)

Nice to have

Hands-on experience building datasets for LLM/VLM/multimodal model pre-training or post-training; Experience with synthetic data, including text generation, image-text synthesis, rendering/diffusion-based generation, or multimodal trajectory generation for GUI, search, file-operation, or agentic tasks; Familiarity with major evaluation benchmarks such as MMLU, MMBench, MM-BrowseComp, with practical experience using evaluation results and failure analysis to drive targeted data improvements; Experience with open file and table formats such as Lance, Iceberg, Paimon and Parquet, including schema evolution, versioning, transactions, indexing, and performance optimization for multimodal AI workloads; Solid understanding of LLMs, speech/audio models, vision models, and multimodal models; Hands-on experience with large-scale AI data construction, cleaning, synthesis, or quality evaluation; Experience building ETL systems, data models, data pipelines, or data warehouses; Experience processing large-scale text, image, or video datasets; Familiarity with Agents and modern LLM toolchains, with practical experience or strong interest in applying LLMs to data production, analysis, quality control, governance, and pipeline automation