Member of Technical Staff
Mundo AI
Member of Technical Staff
Mundo AI
Multilingual multimodal training data provider
Mundo AI builds multimodal data infrastructure for frontier research labs and Fortune 100 companies. Our work covers audio, video, and other new data types that advance perceptual intelligence. We are hiring Members of Technical Staff to create new data infrastructure as well as improve upon existing infrastructure.
We’re looking for an applied researcher to turn open-ended problems into concrete experiments and solutions. You’ll investigate model capabilities and failure modes, develop novel evaluation methodologies, and design datasets that push the boundaries of what models can do. You’ll also work directly with researchers at leading AI labs, translating their research challenges into practical solutions.
This is a high-ownership role. You’ll have the freedom to identify promising research directions, formulate hypotheses, run experiments, and drive projects from initial exploration to real-world impact. We’re particularly interested in people who enjoy both the intellectual challenge of research and the practical work of making things happen.
Investigate model capabilities and limitations. Study how state-of-the-art models behave, identify weaknesses and failure modes, and develop hypotheses about how to improve them.
Design and execute experiments. Turn research questions into rigorous experiments, implement methods, analyze results, and iterate based on findings.
Develop datasets and evaluations. Design novel datasets, benchmarks, and evaluation methodologies that measure model capabilities and inform future improvements.
Work with frontier AI researchers. Collaborate directly with researchers at leading AI labs to understand their challenges, explore potential solutions, and translate research needs into actionable projects.
Explore new research directions. Stay current with advances in AI, identify promising opportunities, and independently propose and pursue research projects.
Turn research into practical outcomes. Go beyond experiments and papers. Translate findings into datasets, evaluations, methodologies, and capabilities that create real value.
Meaningful and relevant experience in applied machine learning, AI research, or research engineering. Exceptional candidates with equivalent research experience are also welcome.
Strong understanding of modern AI models. You’re familiar with contemporary model architectures, training approaches, and inference techniques, particularly in speech, language, or multimodal AI.
Hands-on applied research experience. You’ve done more than prompt models or integrate APIs. You’ve designed experiments, implemented research ideas, evaluated models, and drawn meaningful conclusions from results.
Strong programming skills. You’re proficient in Python and comfortable working with modern machine learning frameworks, research codebases, and experimental workflows.
Research independence. You can take an ambiguous problem, formulate hypotheses, design experiments, and independently drive a research project forward.
Strong analytical and experimental judgment. You understand how to design meaningful experiments, select appropriate metrics, interpret results, and distinguish genuine improvements from noise.
Clear communication. You can explain complex technical concepts, communicate research findings, and collaborate effectively with both technical and non-technical stakeholders.
A strong academic foundation. A bachelor’s degree in computer science, machine learning, statistics, or a related field is expected. A master’s or PhD is preferred, but exceptional research experience can substitute for advanced degrees.
Speech and audio AI expertise. Experience with speech-to-speech models, automatic speech recognition (ASR), text-to-speech (TTS), spoken language understanding, or related areas.
Evaluation and benchmarking experience. You’ve developed benchmarks, designed evaluation methodologies, or investigated model failure modes.
Multimodal AI experience. You’ve worked with models that process or generate audio, video, images, or other modalities.
Model training and post-training experience. You’ve fine-tuned or trained models using techniques such as supervised fine-tuning, reinforcement learning, or preference optimization.
Research infrastructure experience. You’ve built evaluation harnesses, inference pipelines, or tooling to support large-scale experimentation.
Research publications. You’ve published at venues such as NeurIPS, ICML, ICLR, ACL, or Interspeech, or have equivalent demonstrated research contributions.
Experience working with frontier models. You’ve experimented with state-of-the-art open-source or proprietary models and understand their strengths and limitations.
Medical, dental, vision, and mental health coverage
Short- and long-term disability coverage
Health and Lifestyle Spending Accounts
Generous paid time off
Weekly team lunches and wide variety of office snacks and drinks
Regular team events and socials
Mundo AI builds multilingual and multimodal training data and evaluations to help AI labs improve perceptual intelligence. Through end-to-end data operations, it collects and labels audio, text, video, gestures, and real-world interactions in multiple languages using native speakers, delivering proprietary datasets and custom labeling for model training and evaluation. How it works: Mundo runs data collection, annotation, quality control, and evaluation pipelines, creating specialized datasets tailored to clients' needs. Differentiators: it focuses on non-English, multimodal data with fully managed data workflows and native-speaker labeling, backed by a team with experience from AWS, Cohere, Hugging Face, and quant finance, and it has strong funding backing from Y Combinator and venture partners. Goal: to scale high-quality multilingual and multimodal data and evaluation tools so AI models can better understand and interact with the real world across languages and modalities.
Explore more opportunities at Mundo AI