Adrian Cioca

Key Skills

Software

AWS SageMaker

Label Studio

Top Subject Matter

No subject matter listed

Top Data Types

3D Sensor

Top Task Types

Action Recognition

Freelancer Overview

Versatile AI Training Specialist with [X+] years of experience in high-impact data processing for cutting-edge AI systems. Expertise spans Computer Vision (image/video annotation), LLM evaluation, and multimodal data labeling across industries including automotive, healthcare, and conversational AI. Core Skills: Precise annotation for 2D/3D object detection, segmentation & LiDAR data LLM fine-tuning, prompt engineering & bias mitigation Multilingual text annotation ([es: EN, ES, IT]) QA/validation for training datasets Proven ability to handle complex projects with tight deadlines while maintaining >98% accuracy. Passionate about building ethical AI through meticulous data craftsmanship. Open to diverse opportunities: From autonomous vehicles to generative AI, I deliver scalable solutions tailored to your model’s needs.

Labeling Experience

Multimodal AI Training: Autonomous Driving + Conversational AI Dataset

Label StudioTextQuestion AnsweringText Generation

Led annotation for 50K+ multimodal datasets powering self-driving perception systems and multilingual chatbots. CV: Annotated 30K urban scene images/videos (2D bboxes + segmentation) with 98.7% QA accuracy NLP: Labeled 20K user queries for intent classification (EN/ES/IT) across automotive/retail domains Managed 5 annotators, created guidelines, and reduced error rate by 22% via iterative feedback Key outcome: Trained models achieving <0.15% false positive rate in production

Led annotation for 50K+ multimodal datasets powering self-driving perception systems and multilingual chatbots. CV: Annotated 30K urban scene images/videos (2D bboxes + segmentation) with 98.7% QA accuracy NLP: Labeled 20K user queries for intent classification (EN/ES/IT) across automotive/retail domains Managed 5 annotators, created guidelines, and reduced error rate by 22% via iterative feedback Key outcome: Trained models achieving <0.15% false positive rate in production

2025

Education

Work History