Computer Vision, Applied Research Scientist

This is a fully remote position, open to applicants in United States.

📋 Description

• Take ownership of comprehensive experiments on Boon's foundational model, overseeing everything from architecture design to self-supervised pretraining, supervised fine-tuning, and deployment in production.

• Create and assess innovative multi-stage vision architectures aimed at understanding construction drawings.

• Lead decisions regarding backbones, decoders, fusion strategies, loss functions, and training methodologies.

• Conduct thorough experiments incorporating baselines, ablations, and evaluations on authentic construction drawings.

• Explore research avenues to enhance accuracy across various construction trades and scopes.

• Transition models from experimental notebooks to the production inference pipeline.

• Engage hands-on with tools like PyTorch, YOLO, SAM, DINO, and other contemporary computer-vision frameworks.

• Collaborate with ML engineers on deployment strategies, quantization, and serving processes.

• Troubleshoot issues on customer drawings and integrate insights into future training cycles.

• Partner with teams focused on synthetic data, annotation, and infrastructure to address data and compute requirements.

• Collaborate with engineering leadership to shape the accuracy roadmap and strategic direction.

• Draft internal research reports and present findings, trade-offs, and recommendations to engineering leadership.

• Assist in prioritizing data acquisition and annotation efforts.

• Define evaluation datasets and metrics for performance assessment.

• Recognize failure modes and design experiments to tackle them.

⛳️ Requirements

• 3–7+ years of experience in computer vision research.

• Proven track record of published research in computer vision or experience deploying production computer vision models at scale.

• Practical expertise in multi-modal dense prediction, encompassing segmentation, detection, or collaborative vision-language tasks.

• Experience in production with modern vision-transformer backbones such as SAM, DINOv2/v3, CLIP, SigLIP, or similar technologies.

• Strong proficiency in PyTorch and experience in training large-scale vision models.

• Capability to transition models from research environments to production inference pipelines.

• Solid understanding of deep-learning principles, including optimization, loss design, regularization, and self-supervised learning methodologies.

• Proficiency in English, both written and verbal.

• Availability to collaborate during California business hours.

• Experience with Graph Neural Networks or relational reasoning architectures is desirable.

• Familiarity with text spotting, OCR, or scene-text detection is advantageous.

• Experience with LoRA, adapters, or parameter-efficient fine-tuning is a plus.

• Background in self-supervised pretraining is beneficial.

• Experience with engineering or technical drawings, document understanding, or layout analysis is a plus.

• Contributions to open-source computer vision research are appreciated.

• Published works in top venues are an added advantage.

🏝️ Benefits

• Significant equity in an early-stage, well-funded startup.

• Publication-friendly environment — we encourage our team to publish their work at leading venues.

• Autonomy to propose, advocate for, and conduct your own experiments, delivering successful solutions to real customers.

• Collaborate with a top-tier team from esteemed tech companies and research institutions.

• Work at the crossroads of computer vision research and tangible industry impact within a trillion-dollar market.

• Potential for additional compensation, including discretionary bonuses and/or equity in the company.