Secure · Auditable · Human-Controlled Agentic Systems
Building verifiable agentic systems for high-stakes operations.
I build secure, auditable, human-controlled agentic systems for high-stakes operations.
I am the founder of ARKELYTHEX, an engineering venture focused on fiscal-operational software, developer tools, and verifiable automation for Latin America. Its flagship project, Drenyra, explores how AI-assisted fiscal workflows can remain traceable, reviewable, and under human control.
AI assists. The system validates. Professionals control. Evidence remains.
| Area | What I build |
|---|---|
| Evidence-first agentic systems | Agentic workflows designed around explicit state, validation, and inspectable evidence. |
| High-stakes operations | Software for operational domains where control and recovery matter. |
| Developer tools | Tools and environments that improve engineering workflows. |
| Secure automation | Automation designed around isolation, traceability, and human approval. |
| Project | Focus |
|---|---|
| Drenyra / ARKELYTHEX | AI-native fiscal-operational software for LATAM, designed for traceability and human review. |
| DreamFolio | High-performance personal portfolio built with Astro, React, and Tailwind CSS. |
| somnyx-web | Skill-gap analysis platform. |
| somnyx-tui | Terminal client for Somnyx. |
| Dreamcoder Dots | Terminal-first engineering environment for Arch Linux. |
- Build products around real operational constraints, not demos.
- Make authority, state transitions, and validation explicit.
- Use AI within bounded capabilities, never to remove accountable human control.
- Preserve evidence so important results can be inspected, audited, and reproduced.
- Treat security, isolation, and recovery as invariants.
Building the infrastructure that Latin America deserves.