I’m Davi, a systems builder and co-founder of L2 Systems. I build software, AI tools and business automation. My work ranges from helping AI agents carry projects between sessions to sharing data between browsers and exploring rocket designs through simulation.
AI work that continues beyond one conversation.
A workspace that keeps agents’ tasks, tools, project history and approvals together. Built for continuing work across sessions and reviewing what happened.
Public research preview · Windows package · Architecture
Exploring whether the next copy can come from another browser.
A browser-to-browser cache proof of concept, with freshness checks and server fallback. I refined its architecture through discussions with Professor Paulo Coelho at the Federal University of Uberlândia.
Proof of concept · External validation remains in progress.
Computational search for better rocket designs.
Software that generates and evaluates designs for different flight objectives. A fast physics proxy explores candidates; selected designs are checked with OpenRocket.
Experimental software · Simulation and design research.
Through L2, I’m also developing L2 Operations to organize our projects and work. For a dental clinic, I built a website and an AI-assisted content workflow connecting drafts, human review and publication.
I tutored 30+ classmates across core school subjects, creating PDF study guides and later a web study platform shaped by their questions.
How I work with AI
AI is part of my research and development process. I define the problem, make design decisions, direct implementation and review results through tests and iteration. Reusable instructions and project context help me carry work between sessions.
These projects include prototypes and research previews. Their READMEs document the current evidence and limits; collaboration and contributors belong in the project history.
What I’m interested in next
Computer science, distributed systems, applied AI and computational design. I’m interested in how systems behave beyond the first successful demo, and in making the work easier to test, understand and continue.