Md Abu Yusuf

Key Skills

Software

CVAT

Top Subject Matter

No subject matter listed

Top Data Types

No data types listed

Top Label Types

No label types listed

Freelancer Overview

I am an experienced AI Engineer with a strong background in designing and deploying scalable AI systems, with a particular focus on high-quality data pipelines, data curation, and annotation for machine learning and deep learning applications. My expertise spans NLP, computer vision, and multi-agent systems, where I have built and fine-tuned models using Python, PyTorch, TensorFlow, and Hugging Face, ensuring robust data preprocessing, labeling, and validation processes. I have worked extensively with vector databases, data engineering tools like Airflow and Databricks, and implemented ETL workflows to curate and prepare diverse datasets for training AI models in domains such as finance, e-commerce, and mental health. My hands-on experience includes developing retrieval-augmented generation (RAG) pipelines, automating data collection, and applying MLOps best practices to ensure data quality and consistency throughout the AI lifecycle. I am passionate about leveraging my technical skills and industry experience to deliver reliable, well-annotated training data that drives impactful AI solutions.

Labeling Experience

AI Developer / AI Training Contributor

Computer Code ProgrammingEvaluation Rating

Evaluated and annotated LLM-generated outputs, focusing on backend code accuracy and response structure. Reviewed RAG-based chatbot answers for factual consistency and instruction adherence. Provided targeted feedback to improve prompt and model output quality. • Detected reasoning gaps, hallucinations, and structural errors in outputs. • Documented recurring failure patterns to refine datasets. • Supported prompt refinement for logic improvement. • Enhanced overall training data through structured evaluation.

Evaluated and annotated LLM-generated outputs, focusing on backend code accuracy and response structure. Reviewed RAG-based chatbot answers for factual consistency and instruction adherence. Provided targeted feedback to improve prompt and model output quality. • Detected reasoning gaps, hallucinations, and structural errors in outputs. • Documented recurring failure patterns to refine datasets. • Supported prompt refinement for logic improvement. • Enhanced overall training data through structured evaluation.

2025

AI Engineer – Model Evaluation (Omdena)

TextEvaluation Rating

Assessed conversational AI outputs for coherence, bias, and completeness as part of model evaluation initiatives. Conducted comparative testing across multiple LLM providers and performed iterative feedback for model refinement. Curated structured datasets to support consistency in model reasoning. • Evaluated responses for logic and factual accuracy. • Identified bias and hallucination patterns within AI outputs. • Helped improve response quality via feedback loops. • Participated in structured model benchmarking sessions.

Assessed conversational AI outputs for coherence, bias, and completeness as part of model evaluation initiatives. Conducted comparative testing across multiple LLM providers and performed iterative feedback for model refinement. Curated structured datasets to support consistency in model reasoning. • Evaluated responses for logic and factual accuracy. • Identified bias and hallucination patterns within AI outputs. • Helped improve response quality via feedback loops. • Participated in structured model benchmarking sessions.

2025 - 2025

Intern/Thesis – Dataset Validation (Fusion System)

CVATImageObject Detection

Labeled and validated computer vision datasets for pedestrian and cyclist detection projects. Compared AI model predictions to ground truth annotations, performing detailed error analysis. Contributed to dataset quality by supporting ongoing improvement initiatives. • Applied object detection techniques to images. • Analyzed model errors to identify labeling inconsistencies. • Enhanced data integrity for training computer vision models. • Ensured high accuracy in image annotation tasks.

Labeled and validated computer vision datasets for pedestrian and cyclist detection projects. Compared AI model predictions to ground truth annotations, performing detailed error analysis. Contributed to dataset quality by supporting ongoing improvement initiatives. • Applied object detection techniques to images. • Analyzed model errors to identify labeling inconsistencies. • Enhanced data integrity for training computer vision models. • Ensured high accuracy in image annotation tasks.

2023 - 2023

Education

Work History