ARCHITECTING EVALUATION FRAMEWORKS AND TRAINING NEXT-GEN MODELS
Cerveza Patagonia • 2026
Hollywood Production • 2023
Innovation Project • 2024
Creative Direction • 2023
CLICK TO VIEW CASE STUDY
I architect evaluation frameworks and train frontier AI models in creative and multimodal domains. Unlike pure engineers, I bring 20 years of production expertise in Film, VFX, Creative Direction and Digital Marketing, which gives me a unique edge: I understand what "creative quality" actually means, and I can translate that into measurable, inter-rater-reliable rubrics.
20+ years in audiovisual production. From live visuals for electronic music to Hollywood VFX (Mass Effect 3, Halo 4, Cloud Atlas, Twilight) to leading creative teams across YouTube, short-form platforms, and branded media. All og that taught me what "creative quality" actually means.
Now I translate that into measurable, inter-rater-reliable rubrics. I design objective criteria for multimodal evaluation. I red-team failure modes. I optimize RLHF/SFT pipelines. I build scalable QA workflows for generative systems.
Over the years, I've led multidisciplinary teams through the full creative pipeline: concept, scripting, filming, editing, motion graphics, sound design, final delivery. I've spearheaded dozens of digital and video-driven experiences across YouTube, short-form platforms, educational systems, and branded media. Scaled 25+ brands on social using AI-optimized creative workflows. Built 4 enterprise AI video generation playbooks (60% production time reduction). Generated $500K+ in revenue through creative AI integration and strategy. What sets me apart now: I don't just create—I architect systems. I measure quality before production. I design objective rubrics for multimodal evaluation. I red-team failure modes. I optimize RLHF/SFT pipelines. I build scalable QA workflows for generative systems.
The strategic layer matters too. I use data, performance insights, and AI-powered tools to inform decisions, optimize workflows, and scale content production without sacrificing quality or authenticity. My approach is hands-on and collaborative: I work alongside teams, lead by example, and believe great creative work requires clarity, experimentation, and iteration. I'm not a creative guy learning AI. I'm an AI specialist who understands creative quality, production at scale, and how to measure both. That's rare in red-teaming and RLHF spaces.
→ I work with frontier AI companies, research teams, and creative tech organizations on red-teaming, rubric architecture, generative model evaluation, AI training pipelines and Production-scale creative automation.
Goal: Architect evaluation frameworks that ensure frontier AI generates creative outputs of measurable, production-ready quality—enabling teams to scale authentic audiovisual stories and designs with confidence, while empowering communities through accessible creative technology.
If you are interested in a new project, collaboration, or a reason to get a beer don't hesitate to shoot me an email.
hey@thiagotelesfilmmaker.com