Lead analytics and measurement work for Wikimedia's anti-abuse, safety, and security efforts. Responsibilities include designing experiments and metrics, performing quantitative research using Bayesian and frequentist approaches, collaborating with product, research, and engineering teams to develop and evaluate AI/ML models, building self-service dashboards, and working with large-scale data using tools such as Hive, Presto, Druid, Spark, Apache Iceberg, DBT, Airflow, and Superset. Fluency in Python or R, SQL, and experience with experimental design, statistics, and machine learning are emphasized.

NVIDIA seeks a Solutions Architect to work with customers and internal teams to evaluate, benchmark, and improve LLMs, VLMs, embedding models, and agentic pipelines. Responsibilities include driving adoption of NVIDIA evaluation tools (NeMo, Eval Harness, NVIDIA Evaluation Microservice), building proofs-of-concept, analyzing failures and mitigation strategies (model selection, retrieval optimization, fine-tuning), and collaborating with engineering, product, and sales. Requires MS/PhD or equivalent and 5+ years' hands-on experience building and evaluating deep neural networks; strong English communication skills.

Work on inference and serving infrastructure to improve throughput, latency, and KV cache utilization across GPUs and distributed topologies. Responsibilities include kernel-level tuning, quantization without quality regression, profiling and benchmarking, batching and speculative decoding, memory and cold-start optimizations, and setting technical direction. Technology stack includes Python, vLLM, SGLang, TensorRT-LLM, PyTorch, CUDA, Kubernetes and observability/tooling such as Prometheus, Grafana, ClickHouse and GitLab CI.

Design and implement ML models that scale to millions of requests per second, translate business requirements into technical specifications, evaluate and debug models, and provide technical mentorship. Role combines research, engineering, and deployment work in a fully remote, flexible-hours environment. Requires 5+ years professional applied ML experience and experience with real-time, incremental, and online learning.