Erion Barasuol's Work | ContraWork by Erion Barasuol
Erion Barasuol

Erion Barasuol

AI Engineering · Blockchain · Software Development

    2
    Followers
Followed by GALLERY L and Nithish R

São Paulo, Brazil

WorkPostsProductsServicesAbout

São Paulo, Brazil

Cover image for I designed and built Agentic
I designed and built Agentic SDD AutoLearner, a 60-day hands-on learning system that teaches software engineers how to work with AI agents using Spec-Driven Development. The project combines an AI tutoring protocol with executable tooling, persistent learning state, automated verification and phase-based assessments. A core design principle is that the AI acts as the tutor, not the engineer. For most of the program, it can challenge and review a learner’s specifications, but it cannot write them. This forces the learner to develop the engineering judgment required to evaluate AI-generated work later. I designed the system around: • 6 progressive learning phases • 14 executable labs and 14 written drills • 122 automated checks • 6 competency gates with open rubrics • Persistent local progress tracking • Spec-Driven Development workflows • Claude Code integration • A final capstone on a real repository • Zero third-party runtime dependencies The result is a fully local, executable learning system that goes beyond teaching prompts. It trains engineers to specify work, direct AI agents, review their output, verify results and measure the development cycle. Built from concept, pedagogy and architecture through implementation, testing, packaging and commercial release.
0
66
Cover image for A representative architecture for a
A representative architecture for a gold-backed cross-border settlement network, based on my engineering work with proprietary client details removed. I worked across blockchain architecture, technical leadership, backend and distributed systems, designing how a permissioned network could combine reserve-backed issuance, multi-party trust, compliance and auditable settlement. A key architectural decision was separating mutable compliance and personal data from immutable ledger state. Sensitive and erasable information remains off-chain, while the ledger preserves hashes, attestations, supply state and transaction history. The design also enforces reserve-backed issuance at the protocol level: circulating supply cannot exceed independently attested reserves, with multi-party endorsement preventing unilateral issuance. My contribution included: blockchain and trust architecture, endorsement and supply-control design, backend/ledger boundaries, Go chaincode architecture, Hyperledger Fabric integration, and Kubernetes-based deployment design. Stack: Hyperledger Fabric · Go · NestJS · TypeScript · PostgreSQL · Kubernetes · Helm · CI/CD
1
239