Open to opportunities

Murilo da Silva

@murilodasilva

Message

AI Data Trainer and RLHF specialist aligning LLMs with rigorous economic reasoning.

Brazil

Message

What I'm looking for

I’m looking to deepen AI safety impact through RLHF/RLAIF, rigorous LLM evaluation, and red teaming—building golden datasets and prompt systems that reduce hallucinations, bias, and guardrail failures.

I’m an AI Data Trainer and RLHF specialist with a strong economics foundation, focused on building high-fidelity training and evaluation data for safer, more reliable LLM behavior.

I craft golden responses and multi-turn reasoning rationales to improve LLM alignment, with particular emphasis on mitigating hallucinations and bias through rigorous Economic Reasoning and Advanced Prompt Engineering. Since 2024, I’ve consistently run self-directed red teaming and jailbreaking on frontier models to uncover safety vulnerabilities.

In my current contractor role at Mindrift, I author logically structured technical documentation, deliver quality-assured chain-of-thought rationales, and support HITL evaluation aimed at top-tier benchmark performance. I also contribute to AI alignment and red teaming efforts across multiple safety projects (including dataset and bias framework work such as BR-EconEval and LLM-Judge-Econ), applying careful LLM evaluation and adversarial testing to improve guardrails and model validity.

Experience

Work history, roles, and key accomplishments

Current

Performance Analyst & AI Tutor

Current

Mindrift

Jun 2026 - Present (1 month)

Authored high-precision golden responses and complex chain-of-thought rationales to optimize LLM alignment. Produced technically structured documentation and achieved top-tier HITL benchmark scores for quality assurance.

Chain Of Thought LLM Alignment LLM Evaluation Technical Documentation

Current

AI Alignment & Red Team Contributor

Current

Upwork

Jan 2024 - Present (2 years 6 months)

Contributed to AI alignment and safety projects through self-directed adversarial testing and jailbreaking. Identified potential vulnerabilities including prompt injections and guardrail weaknesses.

AI Alignment Red Teaming Jailbreaking Adversarial Testing Prompt Injection Testing

Current

AI Alignment & Red Team Contributor

Current

Aligneer

Jan 2024 - Present (2 years 6 months)

Contributed to AI alignment and safety projects through self-directed adversarial testing and jailbreaking. Identified potential vulnerabilities including prompt injections and guardrail weaknesses.

AI Alignment Red Teaming Jailbreaking Adversarial Testing Prompt Injection Testing

Current

AI Alignment & Red Team Contributor

Current

Meridial Guru

Jan 2024 - Present (2 years 6 months)

Contributed to AI alignment and safety projects through self-directed adversarial testing and jailbreaking. Identified potential vulnerabilities including prompt injections and guardrail weaknesses.

AI Alignment Red Teaming Jailbreaking Adversarial Testing Prompt Injection Testing

AI Contractor & Crowdsourcing

Toloka

Jan 2024 - Jun 2026 (2 years 5 months)

Performed high-volume data curation and search relevance engineering while managing end-to-end digital pipelines. Ensured strict SLA compliance and high data auditability.

Data Curation Search Relevance Engineering Pipelines SLA Data Auditing

AI Contractor & Crowdsourcing

Invisible Technologies

Jan 2024 - Jun 2026 (2 years 5 months)

Performed high-volume data curation and search relevance engineering while managing end-to-end digital pipelines. Ensured strict SLA compliance and high data auditability.

Data Curation Search Relevance Engineering Pipelines SLA Data Auditing

AI Contractor & Crowdsourcing

Clickworker

Jan 2024 - Jun 2026 (2 years 5 months)

Performed high-volume data curation and search relevance engineering while managing end-to-end digital pipelines. Ensured strict SLA compliance and high data auditability.

Data Curation Search Relevance Engineering Pipelines SLA Data Auditing

Education

Degrees, certifications, and relevant coursework

Professional Track (Economic Reasoning & LLM Alignment)

Specialized Professional Track, Economic Reasoning & LLM Alignment

Activities and societies: Self-directed red teaming, jailbreaking, adversarial testing, hallucination/bias mitigation, prompt engineering.

Ongoing specialized training focused on economic reasoning and structural LLM alignment, including approaches to mitigating hallucinations and biases. Includes self-directed red teaming and adversarial testing to identify safety vulnerabilities.