My name is Jady Pamella. I am an AI engineer building production AI systems across LLM platforms, computer vision, evals, and regulated environments, with a focus on AI safety. I came in with twelve years of production engineering across regulated banking, climate tech, and AI products. I am now a PhD Researcher at Aarhus University, working on trustworthy digital twins for critical infrastructure such as water and cooling systems.
I spent eight of those years as IT Manager at Bank of Brasilia in a regulated banking environment, where I rebuilt the DevSecOps release pipeline to meet the compliance requirements of a regulated bank and cut deployment time by 60 percent. That discipline is now anchored on my 2025 paper in Elsevier Computers and Security (CIS-Controls-aligned governance for the Brazilian judiciary) and on the AI engineering work I ship today.
Based in Aarhus, Denmark. I build production AI with a focus on AI safety, and I am open to collaboration on applied AI, LLM products, ML platforms, evals, and AI safety work.
I am building watch-the-watchers, a reproducible evaluation harness for multi-agent collusion and peer-preservation. Current finding: +25 percent in-group protection signal on Gemma 4. When agents judge agents, who watches the watchers?
I am in the BlueDot Impact Technical AI Safety cohort (2026), the EA Sweden Fellowship (2026), and the AGI Strategy course (2026).
I founded SoiQet, an AI-powered social media platform for creators, brands, and teams, built as a multi-tenant LLM platform on Azure with in-house RAG, structured outputs, Stripe billing, Supabase row-level security for tenant isolation, and OAuth across six social networks.
I hold a Master's in Computer and Systems Sciences (AI and IoT) from Stockholm University as a Swedish Institute Scholar, with my thesis at Deep Forestry on real-time LiDAR perception for autonomous below-canopy drone landing (F1 0.926 at 53 ms per frame on Jetson Orin NX, a 29x latency cut).
I also hold a Master of Engineering in Cybersecurity from the University of Brasilia, with research on cybersecurity governance and the CIS Controls framework, which led to the Elsevier paper above.
Three first-place wins at Stockholm AI hackathons (EpiMinds Google AI 2026, Kong AI-Assisted Coding 2025, Kolomolo AI-Assisted Coding 2025). Judge at the Apart Research Global South AI Safety Hackathon (2026), and speaker on the "New Tech, Old Rules" panel at Nordic Tech Week (2026).
I am looking to collaborate on applied AI engineering, LLM products and platforms, computer vision, ML for regulated industries, AI safety evaluations, and audit-grade verification work.
📧 Email me via hello@jadypamella.com
💼 Connect with me on LinkedIn
👀 Check out my GitHub
📝 Read my articles on Medium
🌐 Visit my portfolio at jadypamella.com
📖 Find my publications via ORCID 0009-0006-4280-3364 and Google Scholar
