Darsh Patel

Key Skills

Software

AWS SageMaker

Axiom AI

CVAT

Dataloop

HiveMind

Label Studio

Prodigy

Roboflow

SuperAnnotate

V7 Labs

Internal/Proprietary Tooling

Top Subject Matter

No subject matter listed

Top Data Types

Image

Text

Video

Top Task Types

Bounding Box

Data Collection

Fine Tuning

Object Detection

Tracking

Freelancer Overview

I am a machine learning engineer with 3+ years of experience working hands-on with AI training data across text, image, video, and multimodal systems. My work spans data preparation, annotation design, quality review, and model evaluation for both research and production environments. I have contributed to training and evaluating computer vision models (object detection, tracking, OCR), large language models (reasoning quality, content safety, preference ranking), and multimodal moderation systems. I regularly work on data quality audits, edge case identification, and reviewer-style evaluation to ensure datasets align with real-world deployment requirements.

Labeling Experience

NSFW Images

Label StudioImageBounding BoxClassification

Contributed to multimodal AI systems combining text, image, and video data. Reviewed labeled samples for safety, policy compliance, and real-world robustness, and supported continuous improvement of training datasets used in production ML pipelines.

Contributed to multimodal AI systems combining text, image, and video data. Reviewed labeled samples for safety, policy compliance, and real-world robustness, and supported continuous improvement of training datasets used in production ML pipelines.

2023

Text Moderation (hate speech + gore language)

Internal Proprietary ToolingTextClassificationText Generation

Evaluated and reviewed text datasets for large language models, including reasoning quality, preference ranking, content moderation, and edge case identification. Focused on data correctness, ambiguity resolution, and alignment with expected model behavior.

Evaluated and reviewed text datasets for large language models, including reasoning quality, preference ranking, content moderation, and edge case identification. Focused on data correctness, ambiguity resolution, and alignment with expected model behavior.

2025 - 2025

Worked on image and video datasets involving object detection, tracking, and OCR. Defined annotation guidelines, reviewed bounding boxes and segmentation masks, and performed quality audits to improve dataset consistency for model training and evaluation.

CVATVideoObject DetectionTracking

Using UAV Videos, Small Object Detection and Improving Long Term Trajectory Extraction of Vehicles. Comparing different feature mapping techniques, visualising bounding boxes, identifying anomalies, addressing multi-class overlapping issues, and modifying the SORT algorithm using SIFT triggers.

Using UAV Videos, Small Object Detection and Improving Long Term Trajectory Extraction of Vehicles. Comparing different feature mapping techniques, visualising bounding boxes, identifying anomalies, addressing multi-class overlapping issues, and modifying the SORT algorithm using SIFT triggers.

2022 - 2023

Education

Work History