Role Overview

As an Applied Research intern at Labelbox, you will design, build, and productionize evaluation and post‑training systems for frontier LLMs and multimodal models. You’ll own continuous, high-quality evals and benchmarks (reasoning, code, agent/tool‑use, long‑context, vision‑language, et al.), create and curate post‑training datasets (human + synthetic), and prototype RLHF/RLAIF/RLVR/RM/DPO‑style training loops to measure and improve real‑world task and agent performance.

Your Impact

• Build and own evaluation and benchmark suites for reasoning, code, agents, long‑context, and V/LLMs.
• Create post‑training datasets at scale: design preference/critique pipelines (human + synthetic), and target hard failures surfaced by evals.
• Experiment and prototype RLHF/RLAIF/RLVR/RM/DPO‑style training loops to improve real-world task and agent performance.
• Land research in product: ship improvements into Labelbox workflows, services, and customer‑facing evaluation/quality features; quantify impact with customer and internal metrics.
• Engage with customer research teams: run pilots, co‑design benchmarks, and share practical findings through internal research reports, blog posts, talks, and published papers.

What You Bring

• A strong foundation in AI and machine learning, backed by a Ph.D. or Master’s degree in Computer Science, Machine Learning, AI, or a related field (in progress degrees are acceptable for intern positions).
• A deep understanding of frontier autoregressive and diffusion multimodal models, along with the human and synthetic data strategies needed to optimize them.
• Passion and experience for LLM evaluation and benchmarking.
• Expertise in training data quality construction, measurement and refinement.
• The ability to bridge research and application by interpreting new findings and translating them into functional prototypes.
• A track record of publishing in top-tier AI/ML conferences (e.g., NeurIPS, ICML, ICLR, ACL, EMNLP, NAACL) and contributing to the broader research community.
• Proficiency in Python and experience with deep learning frameworks like PyTorch, JAX, or TensorFlow.
• Exceptional communication and collaboration skills.

Applied Research at Labelbox

At Labelbox Applied Research, we're committed to pushing the boundaries of AI and data-centric machine learning, with a particular focus on advancing human-AI interaction techniques. We believe that high-quality human data and sophisticated human feedback integration methods are key to unlocking the next generation of AI capabilities. Our research team works at the intersection of machine learning, human-computer interaction, and AI ethics to develop innovative solutions that can be practically applied in real-world scenarios.