Open to opportunities

Henrique Lovatti

@henriquelovatti

Message

I evaluate LLM safety, uncover failure modes, and strengthen trustworthy AI behavior.

Brazil

Message

What I'm looking for

I'm looking to evaluate and improve AI systems through LLM safety testing, red teaming, prompt engineering, policy application, and structured risk assessment.

I've evaluated large language model outputs for safety, factual accuracy, alignment, and policy compliance through freelance work and AI task platforms including Outlier and Alignerr.

I assess multi-turn conversations using RLHF annotation frameworks, quality scoring, preference rankings, and structured analytical reports. My work identifies hallucinated sources, fabricated citations, false scientific claims, and refusal failures.

I've designed adversarial red-teaming strategies using jailbreak attempts, context manipulation, prompt injection, and multi-turn escalation. I stress-test models across high-risk safety categories, including bias, hate speech, self-harm, CBRNE, and discriminatory content.

With a Bachelor of Laws and training in prompt engineering, JavaScript, and Python, I apply nuanced policy judgment to distinguish hard refusals, safe redirections, and responses requiring disclaimers.

Experience

Work history, roles, and key accomplishments

AI Red Teaming & Prompt Engineering Specialist

Freelance

Jan 2025 - Dec 2026 (1 year 11 months)

Designed and executed adversarial testing strategies including jailbreak attempts, prompt injection, and multi-turn escalation to evaluate LLM robustness. Produced structured risk assessments with severity classification and mitigation recommendations.

Adversarial Red Teaming Prompt Engineering Jailbreak Testing Risk Assessment LLM Robustness Policy Judgment

AI Evaluator & Trust and Safety Specialist

Freelance

Jan 2024 - Dec 2026 (2 years 11 months)

Evaluated LLM outputs for safety, factual accuracy, alignment, and policy compliance using RLHF annotation frameworks. Applied multi-dimensional quality scoring and identified bias patterns and hallucinated sources.

RLHF LLM Evaluation Trust and Safety Bias Detection Hallucination Detection Quality Scoring Policy Compliance

AI Task Platforms

Outlier, Alignerr

Jan 2024 - Dec 2026 (2 years 11 months)

Completed large-scale AI evaluation and data annotation tasks across multiple platforms with high accuracy and adherence to platform-specific guidelines. Contributed to safety evaluation pipelines and model performance benchmarking projects.

Data Annotation AI Evaluation Safety Evaluation Benchmarking