Marco G.

Key Skills

Software

No software listed

Top Subject Matter

Adversarial AI safety testing and evaluation

Psychological-domain AI training and safety prompt engineering

Adversarial red-teaming and safety rubric development

Top Data Types

Text

Document

Top Task Types

Red Teaming

Prompt + Response Writing (SFT)

Entity (NER) Classification

Freelancer Overview

Handshake AI — Adversarial Red Teamer (Contract). Brings 7+ years of professional experience across legal operations, contract review, compliance, and structured analysis. Core strengths include Internal and Proprietary Tooling. Education includes Master of Science, University of Texas RGV and Bachelor of Science, University of Texas RGV. AI-training focus includes data types such as Computer Code, Programming, and Text and labeling workflows including Red Teaming, Prompt + Response Writing (SFT), and Named Entity Recognition.

Labeling Experience

Handshake AI — Adversarial Red Teamer (Contract)

Red Teaming

Conducted adversarial red-teaming and evaluation of LLM safety, reasoning, and policy compliance using multi-turn attack sequences. Produced and assessed harmful or misuse-oriented prompt variants to identify jailbreak resistance and failure patterns. Translated observed weaknesses into actionable recommendations to expand evaluation coverage and improve safer model behavior. • Designed adversarial prompts and multi-turn attack sequences for safety stress tests • Evaluated frontier model behavior across high-risk edge cases • Identified failure patterns and documented misuse pathway risks • Provided actionable insights to improve evaluation coverage and safety outcomes

2026 - Present

Independent AI Trainer / Evaluator (Contract)

Text

Evaluated LLM outputs for reasoning quality, safety, instruction-following, and edge-case behavior using structured rubrics. Wrote, reviewed, and refined prompts and model responses across diverse evaluation tasks to improve test coverage and assessment quality. Supported dataset improvement and feedback workflows to strengthen overall model performance. • Assessed factuality, nuance, policy alignment, and response quality via rubrics • Authored and refined prompts and evaluation responses across tasks • Rated reasoning quality, safety, and instruction adherence • Contributed to dataset and feedback workflows for model improvements

2025 - Present

Handshake AI — AI Trainer / Prompt Writer / Psychological Domain Expert (Contract)

Prompt + Response Writing (SFT)Entity (NER) Classification

Developed psychology-grounded prompts and datasets to stress-test LLM behavior under safety and persuasion-related conditions. Evaluated responses for accuracy, nuance, and safety edge cases to surface vulnerabilities and risk pathways. Collaborated with AI labs to refine reasoning and safety frameworks used in ongoing model evaluation. • Built prompts and datasets aligned to psychological stress-testing scenarios • Assessed model outputs for accuracy and safety edge-case behavior • Identified vulnerabilities in persuasion, misdiagnosis, and emotional-risk pathways • Partnered with AI labs to improve reasoning and safety frameworks

2025 - Present

Subject Matter Expert, Adversarial Red Teaming - Applause

TextRed Teaming

Subject matter expert supporting adversarial red teaming to assess harmful, biased, and deceptive behaviors in AI systems. Responsibilities included designing adversarial test suites, executing evaluations for reasoning, compliance, hallucination, and privacy risk, and producing safety rubrics for engineering and policy stakeholders. The role required expertise in adversarial prompt engineering, behavioral threat modeling, and applying psychological frameworks to strengthen model alignment. • Designed adversarial test suites to expose harmful or deceptive outputs • Engineered psychologically informed jailbreak and manipulation prompts • Conducted reasoning, compliance, hallucination, and privacy-risk evaluations • Built safety rubrics and threat models to guide mitigations

2024 - Present

Applause — Subject Matter Expert, Adversarial Red Teaming (Contract)

TextRed Teaming

Designed adversarial test suites to expose harmful, biased, deceptive, and noncompliant outputs from AI models. Engineered psychologically informed jailbreaks, manipulation prompts, and obfuscation attacks to probe multiple failure modes. Built behavioral threat models and safety rubrics used by engineering and policy teams, and advised on mitigations to strengthen alignment. • Created adversarial test suites targeting harmful, biased, and deceptive outputs • Engineered jailbreaks and manipulation/obfuscation prompt strategies • Performed evaluations for reasoning, compliance, hallucination, and privacy risk • Produced threat models and safety rubrics to guide mitigations

2024 - Present

Show more

Education

Work History