Shailesh Gaikwad

Key Skills

Software

CVAT

Internal/Proprietary Tooling

Top Subject Matter

No subject matter listed

Top Data Types

Computer Code Programming

Document

Image

Text

Video

Top Label Types

Bounding Box

Computer Programming Coding

Data Collection

Evaluation Rating

Object Detection

Prompt Response Writing SFT

Question Answering

Freelancer Overview

I have over four years of hands-on experience in data validation, annotation QA, and AI/ML model testing, with a strong focus on computer vision and ADAS domains. My work includes large-scale data labeling and quality audits for autonomous driving systems, where I have reviewed and validated more than 25,000 frames for pedestrian, lane, obstacle, and traffic sign detection using tools like CVAT, Labelbox, and Frame Validator. I am skilled in ensuring annotation consistency, defining labeling guidelines, and collaborating with annotation teams to improve dataset quality and standardization. My expertise also extends to validating annotation APIs, preparing QA metrics, and supporting model retraining through detailed error and improvement reports. I am passionate about delivering high-quality, reliable training data that enhances AI model performance, and I am comfortable working with cross-functional teams in fast-paced, Agile environments.As an AI LLM Analyst, I work on GenAI evaluation and content tasks, including comparing text and image-based model outputs, creating and refining prompts, evaluating TTS and vision models, annotating images and videos, segmenting audio, and applying image/video transformations. I ensure data quality, model alignment, and accurate task completion across diverse AI/ML pipelines.

Labeling Experience

AI/LLM Analyst – GenAI QA Tester

Internal Proprietary ToolingImagePrompt Response Writing SFT

As an AI/LLM Analyst, I evaluated AI-generated images and developed reconstruction prompts to enhance visual realism. I designed and refined prompts to guide AI multimodal outputs, ensuring accuracy and consistency in generated media. I conducted quality validation on AI-generated content and optimized prompt design to improve model outputs. • Evaluated and reconstructed interior design images by removing and restoring objects via prompts. • Crafted and iterated image and video prompts for accuracy and fidelity. • Created full conversation scripts to simulate realistic AI interactions. • Collaborated with AI/ML teams in human-in-the-loop QA processes.

As an AI/LLM Analyst, I evaluated AI-generated images and developed reconstruction prompts to enhance visual realism. I designed and refined prompts to guide AI multimodal outputs, ensuring accuracy and consistency in generated media. I conducted quality validation on AI-generated content and optimized prompt design to improve model outputs. • Evaluated and reconstructed interior design images by removing and restoring objects via prompts. • Crafted and iterated image and video prompts for accuracy and fidelity. • Created full conversation scripts to simulate realistic AI interactions. • Collaborated with AI/ML teams in human-in-the-loop QA processes.

2025

LLM Evaluation – AWS Bedrock Chatbot

Aws SagemakerTextEvaluation Rating

I independently developed, tested, and tuned prompts for an LLM-based chatbot using AWS Bedrock. I performed quality evaluation of AI-generated responses for accuracy, tone, and safety. I conducted prompt tuning and analysis to ensure improved user engagement and factual consistency. • Evaluated model responses for hallucinations and context. • Utilized real-time testing interfaces for effective QA. • Assessed chatbot output for appropriateness and safety. • Implemented evaluation metrics for continuous improvement.

I independently developed, tested, and tuned prompts for an LLM-based chatbot using AWS Bedrock. I performed quality evaluation of AI-generated responses for accuracy, tone, and safety. I conducted prompt tuning and analysis to ensure improved user engagement and factual consistency. • Evaluated model responses for hallucinations and context. • Utilized real-time testing interfaces for effective QA. • Assessed chatbot output for appropriateness and safety. • Implemented evaluation metrics for continuous improvement.

2024 - 2025

Data Validation & Annotation Labeling for AI Model Training – ADAS

CVATImageBounding Box

I conducted large-scale data quality and annotation validation for ADAS vision systems, including pedestrian, lane, obstacle, and traffic sign detection. I reviewed and audited over 12,000+ annotated frames for accuracy, bounding box overlaps, and missed labels using CVAT and Labelbox. I ensured consistent, high-quality labeling across multi-camera image datasets and object tracking frames. • Audited annotated frames for class imbalance and tracking drift. • Performed QA for traffic sign classification, lane, and obstacle detection. • Maintained review templates and verification checklists for annotation tasks. • Validated annotation APIs and ensured correct metadata ingestion.

I conducted large-scale data quality and annotation validation for ADAS vision systems, including pedestrian, lane, obstacle, and traffic sign detection. I reviewed and audited over 12,000+ annotated frames for accuracy, bounding box overlaps, and missed labels using CVAT and Labelbox. I ensured consistent, high-quality labeling across multi-camera image datasets and object tracking frames. • Audited annotated frames for class imbalance and tracking drift. • Performed QA for traffic sign classification, lane, and obstacle detection. • Maintained review templates and verification checklists for annotation tasks. • Validated annotation APIs and ensured correct metadata ingestion.

2021 - 2024

SWQA

CVATImageBounding BoxPolygon

Project: AI/ML Data Annotation & QA for Computer Vision Models Pedestrian Detection Model QA QA-reviewed 12,000+ frames used for training pedestrian detection models. Flagged labeling inconsistencies, class errors, and missing boxes. Validated annotation APIs with Postman. ADAS Data Validation QA Tested reverse/parallel parking and lane detection; logged 50+ defects. Lane & Obstacle Detection QA Frame-by-frame validation of lanes, road signs, pedestrians, and obstacles under diverse conditions. Identified false positives/negatives for improved model accuracy. ADAS Annotation & Labeling Audited 25,000+ frames for vehicles, lanes, and pedestrians using CVAT and Labelbox. Ensured sequence consistency and flagged tracking issues. Image data labeling Traffic Sign Classification Dataset QA Reviewed 15,000+ images, ensured class accuracy and format compliance.Labeled and validated 200K+ frames in CVAT, including bounding Created QA checklists and tracked issue resolution in Excel and JIRA.

Project: AI/ML Data Annotation & QA for Computer Vision Models Pedestrian Detection Model QA QA-reviewed 12,000+ frames used for training pedestrian detection models. Flagged labeling inconsistencies, class errors, and missing boxes. Validated annotation APIs with Postman. ADAS Data Validation QA Tested reverse/parallel parking and lane detection; logged 50+ defects. Lane & Obstacle Detection QA Frame-by-frame validation of lanes, road signs, pedestrians, and obstacles under diverse conditions. Identified false positives/negatives for improved model accuracy. ADAS Annotation & Labeling Audited 25,000+ frames for vehicles, lanes, and pedestrians using CVAT and Labelbox. Ensured sequence consistency and flagged tracking issues. Image data labeling Traffic Sign Classification Dataset QA Reviewed 15,000+ images, ensured class accuracy and format compliance.Labeled and validated 200K+ frames in CVAT, including bounding Created QA checklists and tracked issue resolution in Excel and JIRA.

2021 - 2024

Education

Work History