Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

Computer Science student at ETH Zürich with experience building and shipping full-stack AI products at scale. I focus on turning research into practical systems—automating complex pipelines, improving reliability with orchestration and prompt strategies, and measuring impact in real deployments. I’ve worked as a Research Assistant on AI automation and multi-agent content pipelines, built an open-sourced platform as a research artifact, and developed a creator channel reaching 200K+ followers. I’m comfortable across Python/PyTorch and production-ready engineering (APIs, cloud, databases, Docker) and enjoy building end-to-end solutions with meaningful impact.…Computer Science student at ETH Zürich with experience building and shipping full-stack AI products at scale. I focus on turning research into practical systems—automating complex pipelines, improving reliability with orchestration and prompt strategies, and measuring impact in real deployments. I’ve worked as a Research Assistant on AI automation and multi-agent content pipelines, built an open-sourced platform as a research artifact, and developed a creator channel reaching 200K+ followers. I’m comfortable across Python/PyTorch and production-ready engineering (APIs, cloud, databases, Docker) and enjoy building end-to-end solutions with meaningful impact.

Rafael Bragança Oliveira

AI Engineer, Full Stack Developer, Data Scientist, +2





Computer Science student at ETH Zürich with experience building and shipping full-stack AI products at scale. I focus on turning research into practical systems—automating complex pipelines, improving reliability with orchestration and prompt strategies, and measuring impact in real deployments. I’ve worked as a Research Assistant on AI automation and multi-agent content pipelines, built an open-sourced platform as a research artifact, and developed a creator channel reaching 200K+ followers. I’m comfortable across Python/PyTorch and production-ready engineering (APIs, cloud, databases, Docker) and enjoy building end-to-end solutions with meaningful impact.…Computer Science student at ETH Zürich with experience building and shipping full-stack AI products at scale. I focus on turning research into practical systems—automating complex pipelines, improving reliability with orchestration and prompt strategies, and measuring impact in real deployments. I’ve worked as a Research Assistant on AI automation and multi-agent content pipelines, built an open-sourced platform as a research artifact, and developed a creator channel reaching 200K+ followers. I’m comfortable across Python/PyTorch and production-ready engineering (APIs, cloud, databases, Docker) and enjoy building end-to-end solutions with meaningful impact.

Available to hire

Computer Science student at ETH Zürich with experience building and shipping full-stack AI products at scale. I focus on turning research into practical systems—automating complex pipelines, improving reliability with orchestration and prompt strategies, and measuring impact in real deployments.

I’ve worked as a Research Assistant on AI automation and multi-agent content pipelines, built an open-sourced platform as a research artifact, and developed a creator channel reaching 200K+ followers. I’m comfortable across Python/PyTorch and production-ready engineering (APIs, cloud, databases, Docker) and enjoy building end-to-end solutions with meaningful impact.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Intermediate

Intermediate

Intermediate

Intermediate

Language

Portuguese

Fluent

German

Advanced

English

Advanced

Spanish; Castilian

Beginner

Work Experience

Research Assistant (Federated Neural Architecture Search) at Mundo AI Distributed Computing Group, ETH Zürich

March 1, 2026 - September 1, 2026

Developed a communication-aware federated learning method to train one elastic neural network across mobile devices with fluctuating bandwidth. Restored up to ~80% of the accuracy that federated NAS loses on poorly connected clients. Engineered a trace-driven bandwidth simulator and a deadline-truncatable transmission protocol over a nested supernet using PyTorch on a SLURM cluster, sharding oversized updates across clients. Compared Lyapunov control vs. learned (RL/bandit) schedulers for subnet assignment.

Research Assistant (AI Automation) at GenAI Lab, Technical University of Munich (TUM)

October 1, 2025 - March 1, 2026

Built an open-sourced full-stack AI content pipeline platform called InfluenceOS as a practical research artifact. Coordinated complex content generation tasks via structured sub-agent workflows, using Supabase and AI services APIs with state tracking across pipeline phases; reduced manual production time by 80%+. Designed modular orchestration pipelines with n8n and LangGraph to coordinate LLMs and image generation models into a single automated workflow. Developed prompt engineering strategies to improve output consistency across 500+ automated content iterations, enabling reliable pipeline execution at scale.