Open to opportunities

Bruno Paradox

@brunoparadox

I evaluate and train LLMs, delivering accurate, policy-aware QA and multilingual prompt/response judgments for production workflows.

Brazil

Message

What I'm looking for

I’m looking to help teams ship safer, higher-quality AI outputs—through LLM evaluation, RLHF data labeling, and robust prompt/response QA. I value evidence-based rubrics, multilingual judgment, and production-minded workflows.

I’m an AI training and evaluation professional with 16 years of hands-on experience building digital businesses, content systems, and automation workflows. I bring 5+ years of daily, hands-on generative AI work—from OpenAI DaVinci before ChatGPT—and I’ve used LLMs as real production tools.

I focus on prompt/response evaluation, instruction following, and content quality review: comparing answers, detecting hallucinations, assessing safety/policy risk, and writing concise, evidence-based justifications. I build structured rubrics, checklists, and QA-style pipelines that turn ambiguous tasks into repeatable evaluation processes—especially for Portuguese language judgment.

I also bring deep operational experience across AI-assisted content and agent workflows, including multimodal review concepts (text, image, video, and audio). From agent skill systems and knowledge bases to technical QA and tracking logic, my work blends reasoning quality, grounded claims review, and practical domain expertise in marketing, e-commerce, and software/web tasks.

Experience

Work history, roles, and key accomplishments

Current

AI Workflow & LLM Evaluation

Current

Paradox / Nuance Studio

Jan 2025 - Present (1 year 6 months)

Designed AI agent and skill systems for YouTube strategy and built prompt workflows, quality gates, and knowledge bases. Used LLMs for research, summarization, classification, reasoning, and QA/policy checks to evaluate AI outputs for instruction following and quality risks.

AI Agentic Workflow Prompt Engineering LLM Evaluation RLHF Human Feedback Quality Gates Instruction Following Evaluation Knowledge Management Obsidian Markdown YAML TOML Classification Summarization Reasoning

AI-Assisted YouTube Content Ops

Independent

Jan 2022 - Jan 2025 (3 years)

Created and operated AI-assisted YouTube channels using repeatable workflows for research, ideation, scripting, editing, publishing, and performance analysis. Built internal evaluation systems to assess topic/hook fit, narrative structure, retention patterns, and AI-generated content quality including originality, voice, policy risk, and monetization constraints.

Workflows Script Evaluation Hook And Title Evaluation Originality Checks Voice Tone Review Monetization Performance Analysis

Founder, E-commerce & Digital Ops

Own / Family Business

Jan 2013 - Jan 2022 (9 years)

Operated a sporting goods e-commerce business for roughly 10 years, managing catalog, product listings, marketplaces, pricing, customer support, acquisition, and daily operations. Built product content workflows and paid media/funnel operations, and developed AstroWorld, an early OpenAI DaVinci/API automation product connected to Excel and sold via AppSumo.