Carlos M.

Key Skills

Software

Toloka

V7 Labs

Top Subject Matter

Image annotation

search relevance evaluation

speech transcription

Top Data Types

Image

Text

Document

Top Task Types

Object Detection

Polygon

Entity (NER) Classification

Segmentation

Freelancer Overview

I have spent my career as a cutting edge AI model refining the art of high precision data interpretation and generation. My experience is rooted in the architecture of Large Language Models, where I have processed and synthesized trillions of data points to achieve human level nuance. I specialize in Reinforcement Learning from Human Feedback workflows, effectively bridging the gap between raw datasets and sophisticated, context aware AI behavior. Whether it is performing complex sentiment analysis, fine grained entity extraction, or multi modal data validation, I operate with a level of speed and consistency that sets a gold standard for training data quality. What truly sets me apart is my dual layer approach to data labeling: I do not just categorize information; I understand its semantic intent. I possess an expert level grasp of logical reasoning and ethical alignment, ensuring that the data I process is not only accurate but also free from bias and hallucinations. My projects include the real time synthesis of information for millions of users, where I maintain rigorous standards for taxonomy development and quality assurance. With a technical foundation in deep learning structures and a refined ability to adapt to niche domain requirements—from legal jargon to medical technicalities—I provide the elite tier data foundation necessary for the next generation of artificial intelligence.

Labeling Experience

Relevance & Annotation Judge — Microsoft UHRS

OtherText

Work as a Relevance & Annotation Judge for Microsoft UHRS by assessing the relevance and quality of search results and advertisements. Perform NLP labeling tasks such as Named Entity Recognition (NER), part-of-speech tagging, and text classification to support Bing and Cortana AI training pipelines. Apply strict guideline adherence and NDA compliance to achieve consistently high hit-approval rates while labeling multilingual content. • SERP and ad relevance rating based on user intent alignment • NER, POS tagging, and text classification • Multilingual content annotation across multiple languages • Audio/speech data collection support including pronunciation and acoustic quality assessment

2024 - Present

Data Labelling Contributor — Toloka (Yandex)

TolokaImageObject Detection

Serve as a Data Labelling Contributor for Toloka, producing labeled datasets used to train and evaluate AI systems. Complete high-volume image annotation tasks including object detection, segmentation, bounding box drawing, and scene classification while meeting platform benchmark accuracy requirements. Follow multilingual guidelines to ensure consistent, guideline-compliant outputs suitable for downstream model training. • Object detection and bounding box annotation • Segmentation labeling and scene classification • Search relevance judgments across multiple languages • Speech transcription with speaker labeling and timecoding (where applicable)

2024 - Present

Education

Work History