I design and engineer AI systems that bridge industrial applications, machine learning lifecycles, LLM-based architectures, and production infrastructure.

My engineering background spans over 15 years in mission-critical physical systems—including railway traction converters (CAF), early 100% electric transit power systems (JEMA/Irizar), industrial 3D vision robotics (Inser Robótica), and a Doctorate in Electronics & Machine Vision (Tecnun/CEIT).

Today, as AI Architect at Lantek Sheet Metal Solutions, I lead enterprise AI initiatives across CAD/CAM/MES platforms, transforming exploratory algorithms into reproducible, observable, testable, and secure production platforms.

My public technical ecosystem focuses on:

• Reproducible MLOps & LLMOps: Production-packaged pipelines replacing fragile notebooks with versioned data, model lineage, prompt observability, and automated quality gates.
• High-Throughput Agentic Systems: Multi-agent squads, low-latency microservices, and hybrid knowledge graph retrieval (RAG).
• Autonomous Software Factories: Advanced research into Continuous Architecture & Delivery (CA/CD) governed by strict specification contracts and isolated sandboxes.
• Technological Sovereignty: Self-hosted storage infrastructure and SIMD-vectorized cryptographic primitives ensuring zero supply-chain fragility.

┌─────────────────────────────────────────────────────────────────────────────────────────────────┐
│                       CORE IDENTITY: INDUSTRIAL AI & PRODUCTION PLATFORMS                       │
├────────────────────────────────┬────────────────────────────────┬───────────────────────────────┤
│ 1. Industrial AI & Operations  │ 2. MLOps Lifecycle Standards   │ 3. Enterprise LLMOps & Agents │
│    Research & Optimization     │    DVC · MLflow · Hydra · SHAP │    AutoGen · ISO 42010 · RAG  │
├────────────────────────────────┴────────────────────────────────┴───────────────────────────────┤
│                       ADVANCED SPECIALISATIONS: SYSTEMS & AUTONOMY                              │
├─────────────────────────────────────────────────┬───────────────────────────────────────────────┤
│ 4. Autonomous Agent Factories (Dark Gravity)    │ 5. Technological Sovereignty & SIMD Infra     │
│    Hatchet DAG · gVisor Sandboxing · SDD        │    Self-Hosted S3 (MinIO) · AVX-512 Assembly  │
└─────────────────────────────────────────────────┴───────────────────────────────────────────────┘

Bridging physical manufacturing constraints with algorithmic optimization:

• devops_job_shop_scheduling: Mathematical modeling and genetic algorithm heuristic optimization for the Job Shop Scheduling Problem (JSP), minimizing makespan across industrial constraints.
• semantica: Semantic graph modeling, ontologies, and domain-driven knowledge representations for industrial operations.

Packaging machine learning as disciplined software engineering:

• mlops-python-package: Enterprise-grade MLOps foundation implementing reproducible model pipelines (training, evaluation, explanations with SHAP, promotion). Features DVC data/model versioning (dvc.yaml), MLflow experiment tracking, Hydra configuration management, and automated CI test gates.

Orchestrating reliable generative AI and multi-agent coordination:

• llmops-python-package: Domain-driven LLMOps architecture adhering to ISO/IEC/IEEE 42010:2022 and 15289:2019 standards. Provides prompt evaluation, schema-validated agent interfaces, MCP tool integration, and automated OpenWiki architectural documentation.
• fastapi-autogen-team: High-throughput AI service written in native Rust with Axum and Rig, exposing OpenAI-compatible streaming Server-Sent Events (SSE) for multi-agent squads (Planner, Searcher, QA).
• R2R-Engine & R2R-Application: State-of-the-art multimodal Retrieval-Augmented Generation (RAG) platform featuring hybrid knowledge graph indexing and a Next.js/React observability dashboard.

Advanced research in autonomous software delivery under formal governance:

• Dark Gravity (rust_CACD_autonomous_factory): Autonomous Continuous Architecture & Continuous Delivery (CA/CD) factory in Rust. Coordinates a 6-phase DAG via Hatchet, generates Ed25519-signed Non-Human Identity (NHI) verifiable credentials, and executes untrusted agent-written code inside isolated gVisor sandboxes with strict memory clamps.
• speckit-superpowers-bridge: Architectural handoff bridge uniting Spec-Driven Development (Spec Kit—defining WHAT through formal specifications) with Superpowers execution discipline (HOW via TDD, isolated Git worktrees, and subagent review).

Eliminating supply-chain fragility and external licensing risks through sovereign maintenance:

• Sovereign Object Storage Engine: Self-hosted S3 storage tier deployed on Kubernetes (minio, operator, mc, kes, directpv) guaranteeing persistent, independent artifact storage without vendor lock-in.
• Hardware SIMD Vectorization (AVX-512 & ARM64): Maintained and assembly-optimized cryptographic and streaming primitives for wire-speed data pipelines:
  • sha256-simd: Up to 8x hardware acceleration (>3 GB/s per core).
  • highwayhash: Native assembly hashing exceeding 10 GB/s per core.
  • simdjson-go: Gigabit-per-second JSON stream parsing for pipeline telemetry.

• Evidence Before Claims: Every architectural pattern is backed by executable tests, reproducible pipelines, and concrete codebase implementations.
• Human-in-the-Loop (HITL) Gate: Autonomous agents are restricted from merging code into primary branches; production deployments require manual human review and signed gates.
• Zero-Trust Workload Isolation: Sandboxed execution environments with Deny-All egress policies, strict memory clamps, and ephemeral credential lifecycles.
• Architectural Traceability: AST-driven codebase knowledge graphs (graphify), strict ISO/IEC/IEEE 42010 viewpoint mapping, and automated living documentation.