Peter Robicheaux

ML Lead at Roboflow

Ink-wash portrait of Peter Robicheaux

Papers and workshops

ICLR 2026

  • First real-time detector past 60 AP on COCO
  • Over 380 citations

RF-DETR: Neural Architecture Search for Real-Time Detection Transformers

Isaac Robinson, Peter Robicheaux, Matvei Popov, Deva Ramanan, Neehar Peri

Real-time detection transformer on a DINOv2 backbone. Instead of hand-picking an architecture, fine-tune once and evaluate thousands of speed/accuracy trade-offs without retraining. Does segmentation too.

NeurIPS 2025 D&B

Roboflow100-VL: A Multi-Domain Object Detection Benchmark for Vision-Language Models

Peter Robicheaux*, Matvei Popov*, Anish Madan, Isaac Robinson, Joseph Nelson, Deva Ramanan, Neehar Peri. *Equal contribution.

100 detection datasets of things VLMs rarely saw in pre-training. Strong zero-shot models get under 2% on the hard medical sets. The goal: teach a model a new concept from a few examples and written instructions.

CVPR workshops, 2025 and 2026

Foundational Few-Shot Object Detection Challenge

Helped run the challenge with collaborators at Carnegie Mellon. The 2025 winners beat our baseline by 17 mAP.

Platform

Model training

Bringing new architectures into Roboflow training as they come out: YOLOv8 instance segmentation, YOLO11, YOLOv12, YOLO26, and RF-DETR from the first release through the larger sizes and segmentation.

VLM fine-tuning

PaliGemma 2 and Florence-2 fine-tuning with LoRA, plus the training and deployment path behind it. Cosmos 3 Edge after that, on images and video.

Training pipeline

The parts every job shares: early stopping, automatic batch sizing for RF-DETR, out-of-memory fixes, and training more than one model on the same dataset version.

Experiment platform

Built the research team’s internal experiment platform. Launch training runs on our GPU cluster from a CLI, fan a benchmark like Roboflow100-VL out to a hundred jobs, and keep results and artifacts in one place.

Anomaly detection

PatchCore and FoundAD. Only needs images of good parts, and gives you heatmaps.

Action recognition

Train action recognition on your own video in Roboflow. Label time ranges, fine-tune Cosmos 3 Edge or V-JEPA 2.1, then deploy to the cloud or your own GPUs.

Architecture search

The search from the RF-DETR paper. Run it on your own dataset, for your own hardware.

Serving

Roboflow Inference, the open-source server. I wrote the async model manager and added serving for PaliGemma, Florence-2, SAM 2, OWLv2, Perception Encoder, a lot of YOLOs, RF-DETR and the anomaly models.

Segment Anything

Worked on the SAM 2 backend for Smart Polygon in the labeling tool, including multi-polygon masks. Stood up the SAM 3 video tracker behind click-to-track video annotation.

Roboflow Labs

Contributing to Nebula, the contributor platform behind Roboflow Labs. Building inference infrastructure for AI-assisted labeling and leading a data project.

Hiring and partnerships

Hiring

Building a team of incredibly cracked MLEs and researchers.

Partnerships

Part of Roboflow’s partnership work with NVIDIA, through the Cosmos Coalition, and with other companies. Brought Cosmos 3 post-training to Roboflow.

Talks

Open source