AI & Enterprise Architect specialized in designing scalable, cloud-native and agentic AI architectures — bridging business strategy, emerging AI technology, and enterprise execution.
At BBVA I lead the definition of the bank's AI & Cloud architecture strategy — sitting at the intersection of business strategy, technology architecture, and innovation, not just implementation. 13+ years moving from software developer → IT architect → architecture manager → AI & Enterprise Architecture lead gives me both the boardroom fluency to align AI initiatives with business goals and the hands-on depth to prototype what I propose before asking a team to build it. Below is my public lab: 10 projects spanning agentic banking infrastructure (ledgers, payments, contact-center AI), the AI platform layer underneath them (agent orchestration, observability, inference), and the tooling I built to get there — all self-hosted, no vendor lock-in. Most of my enterprise architecture work at BBVA itself is confidential.
| Stage | Focus |
|---|---|
| Now — AI & Enterprise Architecture Lead | AI strategy, agentic & cloud-native architecture, business-technology alignment, team leadership |
| Manager of Architects | Core banking, service architecture, enterprise solutions across business units |
| IT Architect | SOA, microservices, FaaS/serverless, cloud computing |
| Software Developer (foundation) | Java, C#, PHP, Python — the hands-on base my architecture decisions still draw on |
🏦 Agentic Banking & Fintech
| Project | What it does |
|---|---|
| aletheia-call-agent | Real-time assist for bank contact centers (es-PE): streaming ASR, emotion detection, merchant-descriptor resolution, dispute classification under Peru's Ley 31763 — rules classify, the LLM only justifies. |
| aerarium-agentic-banking | Banking core for agentic commerce — a real Rust append-only double-entry ledger in micro-units, hold→capture authorizations, consent-backed payment mandates. |
| mercatus-agentic-payments | Dual-sided framework for agentic commerce — AI agents that discover, pay for, and monetize services, with swappable x402 (USDC) and AP2 (mandate-based) settlement adapters. |
| veritium-document-intelligence | Agentic document extraction/validation: hexagonal architecture, bounded ReAct loop, rule-based validation with evidence grounding (bbox + page per field), full OTel tracing. |
🧠 Agent & AI Platform Infrastructure
| Project | What it does |
|---|---|
| aeon-agent-harness | Self-hosted agent harness (early scaffolding): Go control plane with Cedar authz and multi-provider model gateway, durable Temporal execution, interop with LangGraph, CrewAI, OpenAI Agents SDK, Microsoft Agent Framework, and Claude Agent SDK. |
| argus-observability-platform | OTel-native observability & agentic AIOps for AI systems — GenAI cost/token attribution, sub-2s hot-path anomaly detection. Running as a live pilot against my own infra (68k+ real spans, 229 tests). |
| prometheus-inference-platform | Self-hosted LLM inference gateway — JWT auth with per-model scopes, multi-backend routing (llama.cpp/MLX/vLLM/SGLang), the infra backbone several projects above run on. |
🛠️ AI Tooling & Frameworks
| Project | What it does |
|---|---|
| synaptum-framework | Minimal, provider-agnostic durable agent runtime — generator-based execution with checkpoint/resume and deterministic step identity, zero required dependencies. |
| axonium-sdk | Python, Go and Rust client SDKs for my Prometheus inference gateway — OAuth2 token rotation, typed error handling, SSE streaming, rate-limit visibility. |
| prosodia-media-platform | 100% local, open-source video dubbing (EN→ES): Whisper transcription, pyannote diarization, context-aware LLM translation, per-speaker voice cloning. No paid APIs. |
(Pinned on my profile — check the repo list for earlier work too.)
Building hands-on — evidenced in the projects above
Applying at architecture & leadership level — real professional experience, not shown in public code
Foundation — earlier career
- 🏛️ Enterprise Architecture — TOGAF-aligned strategy, bridging technology and business goals across a regulated financial institution.
- 🧠 AI & Agentic Architecture — designing agentic systems, RAG/hybrid retrieval, multi-agent patterns, and LLM integration strategy — from proof-of-concept to enterprise rollout.
- 🏦 Agentic Commerce & Fintech Infrastructure — ledgers, payment mandates, and compliant contact-center AI for regulated finance — durable orchestration (Temporal) as a first-class primitive.
- ☁️ Cloud Strategy — multi-cloud architecture (AWS/Azure/GCP), cost optimization, and cloud-native/serverless design.
- 🔗 Integration Patterns — SOA, EDA, Event Sourcing, EIP for complex, high-availability enterprise systems.
- 🖥️ Hands-on LLM infrastructure — self-hosted inference gateways, multi-backend routing, quantization (Q4/Q5/Q6) — demonstrated in shipped open-source projects.
- 🔐 Security & Compliance — secure architectures for regulated industries: JWT/RBAC, PII masking, guardrails, cryptography.
- 👥 Technical Leadership — building and mentoring the team that defines BBVA's AI & Cloud architecture strategy.
Already applying, going deeper on:
- Agentic commerce infrastructure for regulated finance — durable agent orchestration (Temporal), append-only ledgers, and consent-backed payment mandates as first-class architecture primitives, not add-ons.
- Multi-agent orchestration at enterprise scale, and what "AI-native" enterprise architecture looks like when agents — not services — are the unit of design.
- Small language models & edge/local inference as a deliberate strategy, not just a cost hack.
Actively studying — genuine gaps I'm closing, relevant to banking/regulated AI:
- Agent identity, governance & audit-trail frameworks for autonomous systems.
- EU AI Act & DORA compliance requirements for agentic architectures in financial services.
- FinOps for AI — GPU/token cost governance at production scale.
- Context engineering as a discipline distinct from RAG.
- Sovereign AI & data-residency patterns for regulated, multi-region deployments.
Continuous learning, open-source AI tooling, and mentoring engineers moving from development into architecture.