Vishesh Sompura

Key Skills

Software

CVAT

Roboflow

Internal/Proprietary Tooling

Top Subject Matter

No subject matter listed

Top Data Types

Geospatial Tiled Imagery

Image

Video

Top Label Types

Bounding Box

Classification

Object Detection

Polygon

Segmentation

Freelancer Overview

I’ve worked hands-on with AI training data across marine safety, defence, and agriculture, owning everything from label schema design to quality control. For SharkSpotter 2.0, I defined the annotation guidelines for aerial imagery (sharks, swimmers, vessels, surf conditions), created clear instructions for annotators, and set up multi-stage review processes to catch inconsistencies and edge cases. I also worked closely with domain experts to refine labels over time so the dataset stayed aligned with real-world operating conditions. In AgTech and defence projects, I’ve curated and labeled large volumes of RGB and multispectral drone imagery for crop-health mapping, as well as sensor data for autonomous systems. This included designing class ontologies, managing dataset versions, running systematic spot-checks on annotations, and using model feedback to identify label gaps or drift. Across these projects, my focus has been on building reliable, well-documented datasets that directly support robust, repeatable AI model training.

Labeling Experience

AgTech Drone Imagery - Crop Health & Disease Mapping

Internal Proprietary ToolingGeospatial Tiled ImageryPolygonGeocoding

Curated and labeled drone imagery over farms to map crop health, disease patches, and field boundaries. Drew polygons around stressed vs healthy vegetation, aligned labels with NDVI/NDRE layers, and tagged plots across different growth stages and seasons. Implemented systematic spot-checks and versioned datasets as labeling guidelines evolved.

Curated and labeled drone imagery over farms to map crop health, disease patches, and field boundaries. Drew polygons around stressed vs healthy vegetation, aligned labels with NDVI/NDRE layers, and tagged plots across different growth stages and seasons. Implemented systematic spot-checks and versioned datasets as labeling guidelines evolved.

2024 - 2025

SharkSpotter 2.0 - Aerial Marine Safety Imagery Labeling

CVATImageBounding BoxObject Detection

Labeled large volumes of drone imagery of beaches to identify sharks, swimmers, vessels, flags, and surf zones. Designed the labeling schema and detailed guidelines, then performed and reviewed annotations for consistency across edge cases (glare, turbidity, occlusions). Managed multi-stage quality checks with domain experts to keep labels aligned with real-world lifeguard operations.

Labeled large volumes of drone imagery of beaches to identify sharks, swimmers, vessels, flags, and surf zones. Designed the labeling schema and detailed guidelines, then performed and reviewed annotations for consistency across edge cases (glare, turbidity, occlusions). Managed multi-stage quality checks with domain experts to keep labels aligned with real-world lifeguard operations.

2024 - 2024

Defence Autonomous Systems - Sensor & Vision Dataset Labeling

Internal Proprietary ToolingImageBounding BoxTracking

Labeled multi-sensor data from autonomous vehicle trials, including obstacles, waypoints, and mission-relevant zones in camera frames. Assigned persistent IDs for tracked objects across sequences and tagged events such as near-collisions, path deviations, and environment changes. Established QA checklists and sampled batches for re-review to keep labels consistent over long-duration missions.

Labeled multi-sensor data from autonomous vehicle trials, including obstacles, waypoints, and mission-relevant zones in camera frames. Assigned persistent IDs for tracked objects across sequences and tagged events such as near-collisions, path deviations, and environment changes. Established QA checklists and sampled batches for re-review to keep labels consistent over long-duration missions.

2022 - 2022

Education

Work History