CoreFrame Labs | AI Agricultural Robotics

About CoreFrame Labs

Robotics & AI innovation — building practical automation for real-world industries.

CoreFrame Labs is a robotics and artificial intelligence company focused on developing compact, reliable automation systems. We design intelligent machines powered by computer vision, edge AI, and real-time robotics engineering.

Our first flagship project, SCAAR, is an agricultural perception system for future robotic integration. The released SCAAR-Perception-v1.0 baseline achieves 94.96% mAP@0.50 and 95.64% recall on a held-out, family-isolated strawberry benchmark. These results measure software perception, not physical harvesting performance.

Our goal is to validate this concept, demonstrate real-world capability, and expand CoreFrame’s robotics platform into additional industries that benefit from intelligent automation.

✅ Edge-optimized AI vision using YOLO architecture
✅ Early robotic arm concept for gentle, low-damage fruit picking
✅ Developing an integrated AI-robotics pipeline for future field deployment
✅ Roadmap toward autonomous robotics across multiple industries

Technology Stack

Built on modern AI, robotics, and edge compute.

Computer Vision

Ripe, unripe, and rotten strawberry detection using YOLO-based computer vision models trained on curated strawberry datasets with independent CoreFrame field imagery used for validation and ongoing development. Optimized for real-time inference in field environments.

YOLO-based detection with experimental secondary berry-vs-leaf validation and temporal tracking under development

Robotics

Modular robotic grabber concept designed for gentle, low-damage fruit handling. Built to adapt to varying field layouts, plant spacing, and mounting configurations.

Soft-contact grabber tool for delicate fruit handling

Edge Hardware

Targeted deployment on edge AI hardware platforms with integrated cameras and sensors, enabling low-latency autonomous operation without cloud dependency.

Jetson-class edge AI deployment target

SCAAR STRAWBERRY PERCEPTION

Multi-stage computer vision for strawberry detection, ripeness classification, foliage rejection, and persistent tracking.

94.96% mAP@0.50 Released baseline benchmark

Recall: 95.64%
F1: 91.79%
Matched-Object Classification Accuracy: 98.82%

Earlier SCAAR prototype field demonstration.
Shown for visual context; current benchmark results are measured separately.

Released Baseline

SCAAR-Perception-v1.0

556-image family-isolated benchmark
1,928 labeled fruit
Ripe / Unripe / Rotten
Released perception baseline

Current R&D

Experimental · Unreleased

84% fewer foliage false detections
107 → 17 on a separate 171-image internal hard-negative holdout
99.08% validator accuracy
100% berry recall on a 541-crop held-out validation test
Tracking + Re-Identification
Persistent fruit identity remains under active development

Software perception results only. Physical harvesting performance is evaluated separately.

Perception-to-Targeting Validation

Earlier SCAAR prototype testing demonstrated conversion of AI-generated strawberry detections into simulated targeting coordinates under outdoor agricultural conditions.

Historical Prototype Capabilities

Simulated robotic targeting path generation
Real-time coordinate extraction from AI detections
Multi-target detection under outdoor field conditions
Early perception-to-action feasibility validation
~190–235 ms — Historical prototype latency on earlier hardware/software configuration

Since this earlier prototype, CoreFrame Labs has improved the underlying perception stack through family-isolated benchmarking and experimental hard-negative foliage training and secondary candidate validation. Persistent object tracking and identity persistence remain under development.

2D targeting simulation only — not physical robotic actuation

Robotic Harvesting System Concepts

Early-stage robotic harvesting concepts connecting perception, motion, and gentle fruit handling into a focused Phase I development architecture.

CoreFrame Labs is developing a Phase I robotic harvesting architecture designed to connect AI-powered strawberry detection with lightweight robotic movement and gentle fruit handling.

The SCAAR arm concepts shown below represent early-stage engineering direction for feasibility validation, prototype planning, and perception-to-action integration. These visuals are conceptual design studies, not final production hardware.

Key Focus Areas

Compact 4-DOF harvesting arm architecture
Short-range fruit targeting and transfer
Soft-contact end-effector design
Lightweight modular construction
Field-focused prototype iteration

Live conceptual CAD animation showing the Phase I harvesting loop from approach through transfer.

Research & Industry Analysis

Industry research supporting CoreFrame Labs' work in agricultural automation and AI-powered harvesting systems.

Market Research

Market Research on Strawberry Harvesting and Automation

Strawberry farming is one of the most labor-intensive sectors in agriculture, with harvesting requiring up to 3,000 labor hours per acre. The crop must be picked repeatedly by hand throughout the season and relies on large seasonal harvest crews. This research examines production scale, labor demands, and the growing opportunity for AI-assisted automation in specialty crop agriculture.

Sources: USDA NASS, USDA ERS, UC Davis, University of Florida IFAS, Texas A&M AgriLife

Economic Analysis of Strawberry Harvesting and AI-Assisted Automation

Harvest and postharvest operations can represent up to 70% of total strawberry production costs in major U.S. growing regions, with per-acre harvest costs exceeding $79,000 in some California production systems. This research evaluates the economic structure of harvesting, cost drivers for growers, and how AI-assisted systems could improve efficiency and reduce labor costs.

Sources: USDA NASS, USDA ERS, UC Davis, UF IFAS, Texas A&M AgriLife, U.S. Department of Labor

Meet CoreFrame Labs

Leadership driving CoreFrame Labs' AI and robotics mission from strategy to deployment.

Dakotah Singh

Co-Founder & Chief Technology Officer

Dakotah leads the design and deployment of end-to-end AI and robotics systems at CoreFrame Labs. He specializes in applied machine learning, computer vision, and real-time inference, focusing on building production-oriented systems designed for real-world deployment.

He architected and developed CoreFrame Labs’ agricultural perception stack, progressing from early prototypes to the released SCAAR-Perception-v1.0 baseline achieving 94.96% mAP@0.50 and 95.64% recall on a held-out, family-isolated strawberry benchmark. These are software perception results, not physical harvesting results. His work emphasizes continuous model iteration, dataset engineering, and performance optimization to improve reliability in real-world conditions.

Dakotah designs scalable ML pipelines and data processing workflows, overseeing the full AI lifecycle from data collection and labeling to training, optimization, and deployment. He also leads the integration of AI systems with robotic hardware, developing real-time perception for future field automation.

Jessica Singh

Co-Founder & Chief Executive Officer

Jessica brings over 25 years of experience in business leadership, including 15 years in the technology sector, with a strong background in operations, finance, and strategic execution. As CEO of CoreFrame Labs, she leads the company’s strategic direction, partnerships, and financial oversight, ensuring the successful transition of advanced AI and robotics technologies into real-world applications.

She has a proven track record of managing and scaling business and technology-driven operations, with a focus on execution, sustainability, and long-term growth. At CoreFrame Labs, she is responsible for building industry relationships, securing strategic opportunities, and aligning the company’s AI and robotics innovation with real market demand.

Contact & Partnerships

Interested in pilots, grants, or joint development?

AI & Robotics for Real-World Automation