Antonio Menchca

Key Skills

Software

Axiom AI

Clickworker

CloudFactory

CrowdSource

CVAT

Datature

Doccano

Figure Eight

Google Cloud Vertex AI

Hasty

Kili Technology

LabelImg

Mindrift

Snorkel AI

VoTT

Internal/Proprietary Tooling

Data Annotation Tech

Top Subject Matter

No subject matter listed

Top Data Types

3D Sensor

Computer Code Programming

Image

Top Task Types

Bounding Box

Computer Programming Coding

Emotion Recognition

Fine Tuning

Geocoding

Freelancer Overview

Expert AI Trainer and Data Annotation Specialist who has worked on projects related to multimodal dataset (text, image, audio, video) curation to evaluate LLM, NLP, computer vision, and speech recognition with 5+ years of experience. Experienced in feedback-driven processes, model-aided labeling, active learning cycles, and enhanced data quality and throughput by 30%. Skilled in Clarifais owned platform, Labelbox, CVAT, Prodigy, Doccano, SuperAnnotate and self-created research pipelines. Researcher in NLP (NER, sentiment tagging, POS tagging, classification), computer vision (object detection, segmentation, image classification), and ASR transcription. Well versed in TensorFlow, PyTorch, Keras, Scikit-learn, OpenCV, and spaCy as well as preprocessing (tokenization, normalization, augmentation) and QA. Familiar with model assessment on the accuracy, precision, recall, F1, mAP, ROC-AUC, and cross-validation. Works with engineers and researchers to develop guidelines, tools to refine and scalable and consistent annotation. Healthcare, finance, autonomous systems, retail analytics, and generative AI projects (GANs, VAEs, transformers) are all considered industry exposure. AMC loves ethical AI, mitigating bias, and creating datasets that drive more capable and safer machine learning systems.

Labeling Experience

Multilingual Speech Transcription & Audio Annotation

Data Annotation TechVideoEntity Ner ClassificationEmotion Recognition

Transcribed and labeled speech datasets in multiple languages to support training of automatic speech recognition (ASR) and voice assistant systems. Annotated speaker turns, background noise, intent categories, and emotion states to improve natural language understanding and contextual responses. Performed time-aligned segmentation to create precise training examples for model training. Implemented multi-pass QA and peer review to ensure transcription accuracy above 98%. Contributed to improved ASR model accuracy across diverse accents and noisy environments.

Transcribed and labeled speech datasets in multiple languages to support training of automatic speech recognition (ASR) and voice assistant systems. Annotated speaker turns, background noise, intent categories, and emotion states to improve natural language understanding and contextual responses. Performed time-aligned segmentation to create precise training examples for model training. Implemented multi-pass QA and peer review to ensure transcription accuracy above 98%. Contributed to improved ASR model accuracy across diverse accents and noisy environments.

2023 - 2023

LLM Evaluation & Generative AI Dataset Curation

DoccanoTextEntity Ner Classification

Curated, labeled, and evaluated large-scale text datasets for training and fine-tuning transformer-based language models. Performed structured annotation tasks including named entity recognition (NER), intent classification, summarization, toxicity/bias tagging, and human preference ranking for reinforcement learning from human feedback (RLHF). Developed and refined annotation guidelines, participated in iterative feedback cycles with machine learning engineers, and implemented double-pass QA reviews to ensure consistency and high inter-annotator agreement. Supported red-teaming efforts to surface unsafe outputs and enhance model safety alignment. This project contributed directly to measurable improvements in model performance, reducing harmful outputs and increasing user satisfaction.

Curated, labeled, and evaluated large-scale text datasets for training and fine-tuning transformer-based language models. Performed structured annotation tasks including named entity recognition (NER), intent classification, summarization, toxicity/bias tagging, and human preference ranking for reinforcement learning from human feedback (RLHF). Developed and refined annotation guidelines, participated in iterative feedback cycles with machine learning engineers, and implemented double-pass QA reviews to ensure consistency and high inter-annotator agreement. Supported red-teaming efforts to surface unsafe outputs and enhance model safety alignment. This project contributed directly to measurable improvements in model performance, reducing harmful outputs and increasing user satisfaction.

2023 - 2023

Computer Vision Image & Video Annotation for Object Detection

CVATImageBounding BoxPolygon

Annotated thousands of images and video frames for object detection, classification, and semantic segmentation tasks. Applied pixel-accurate bounding boxes, polygons, and masks to label objects including vehicles, pedestrians, road signs, retail products, and manufacturing defects. Conducted frame-by-frame tracking for multi-object scenarios to support video-based model training. Developed annotation guidelines to ensure consistency across a distributed team of annotators and performed double-pass quality checks to achieve >95% inter-annotator agreement. Delivered datasets that were production-ready and contributed to improved model accuracy and real-world robustness.

Annotated thousands of images and video frames for object detection, classification, and semantic segmentation tasks. Applied pixel-accurate bounding boxes, polygons, and masks to label objects including vehicles, pedestrians, road signs, retail products, and manufacturing defects. Conducted frame-by-frame tracking for multi-object scenarios to support video-based model training. Developed annotation guidelines to ensure consistency across a distributed team of annotators and performed double-pass quality checks to achieve >95% inter-annotator agreement. Delivered datasets that were production-ready and contributed to improved model accuracy and real-world robustness.

2022 - 2023

Education

Work History