Nestor Omar Almanza Morales

Key Skills

Software

CVAT

Labelbox

Argilla

Top Subject Matter

No subject matter listed

Top Data Types

Image

Video

Top Label Types

Evaluation Rating

Red Teaming

RLHF

Freelancer Overview

I am an experienced AI trainer and red-teaming specialist with a strong focus on data labeling, annotation, and human-in-the-loop verification for advanced generative models. My expertise spans RLHF, adversarial testing, and high-volume multimodal data evaluation, including hands-on work with Google Veo, Kling, and Luma. I have led benchmarking projects involving over 500 text-to-video model evaluations, meticulously documenting temporal consistency, safety, and physics-based rendering. My background also includes managing biometric data annotation for facial recognition systems, ensuring ground-truth accuracy and technical robustness. I leverage tools such as DaVinci Resolve for aesthetic assessment, n8n for automation, and custom APIs for data validation, with experience in both visual content creation and large-scale operational environments. My commitment to precision and security in AI training data is demonstrated through my work in biometric validation, cinematic quality standards, and real-world logistics for high-stakes events.

Labeling Experience

AI Pipeline Evaluator – Masterize.tech

LabelboxVideoRed Teaming

As an AI Pipeline Evaluator, I executed over 1,000 generative video model evaluations. I stress-tested outputs from advanced platforms including Google Veo and Kling for temporal coherence and adherence to ethical/physical constraints. My role included engineering adversarial datasets to optimize safe deployment of generative video outputs. • Performed complex red-teaming audits on video generative models. • Identified and documented motion artifacts and hallucinations. • Developed datasets targeting model weaknesses. • Collaborated on multi-modal benchmarks for deployment readiness.

As an AI Pipeline Evaluator, I executed over 1,000 generative video model evaluations. I stress-tested outputs from advanced platforms including Google Veo and Kling for temporal coherence and adherence to ethical/physical constraints. My role included engineering adversarial datasets to optimize safe deployment of generative video outputs. • Performed complex red-teaming audits on video generative models. • Identified and documented motion artifacts and hallucinations. • Developed datasets targeting model weaknesses. • Collaborated on multi-modal benchmarks for deployment readiness.

2025

Data Validation Lead – SICOAS (AI Biometric)

CVATImageEvaluation Rating

As Data Validation Lead for an AI biometric project, I managed the manual annotation and verification of over 5,000 biometric data points. My work focused on delivering high-fidelity ground-truth labeling with zero-margin error for facial recognition training datasets. I also conducted critical accuracy audits for system modules before national-level showcases. • Led annotation and validation initiatives for facial recognition. • Implemented strict QA protocols for ground-truth dataset reliability. • Coordinated module reviews in advance of major demos. • Ensured compliance with national and industry accuracy standards.

As Data Validation Lead for an AI biometric project, I managed the manual annotation and verification of over 5,000 biometric data points. My work focused on delivering high-fidelity ground-truth labeling with zero-margin error for facial recognition training datasets. I also conducted critical accuracy audits for system modules before national-level showcases. • Led annotation and validation initiatives for facial recognition. • Implemented strict QA protocols for ground-truth dataset reliability. • Coordinated module reviews in advance of major demos. • Ensured compliance with national and industry accuracy standards.

2024

Qualitative Evaluator – Camp Medolark

ArgillaImageRLHF

During a specialized summer program as Qualitative Evaluator, I contributed human-in-the-loop feedback for over 1,000 creative training samples. I applied strict aesthetic and quality criteria for high-fidelity reinforcement learning signals in creative projects. This input enhanced RLHF-based model training feedback loops for image generation. • Provided expert benchmark feedback for aesthetic model training. • Maintained consistent annotation standards across projects. • Integrated user-oriented quality audits. • Supported iterative improvements through repeated evaluation cycles.

During a specialized summer program as Qualitative Evaluator, I contributed human-in-the-loop feedback for over 1,000 creative training samples. I applied strict aesthetic and quality criteria for high-fidelity reinforcement learning signals in creative projects. This input enhanced RLHF-based model training feedback loops for image generation. • Provided expert benchmark feedback for aesthetic model training. • Maintained consistent annotation standards across projects. • Integrated user-oriented quality audits. • Supported iterative improvements through repeated evaluation cycles.

2025 - 2025

Education

Work History