Álvaro Arcos

I build LLM agents, RAG pipelines, and computer-vision systems that run in production. CTO & AI Director at MCCM Innovations.

It all started with road signs: my PhD work ranked #1 on the German Traffic Sign Recognition Benchmark. Fifteen years later, my teams ship agents and perception systems for enterprises.

Portrait of Álvaro Arcos — → returned me.webp · 42 ms

S01Profile

mixture_of_experts: builder · leader · researcher

Since 2018 I lead engineering and AI at MCCM Innovations, where my teams ship agentic systems, document intelligence, and computer-vision products for enterprise clients across logistics, finance, health, and energy. The common thread of my career: taking models out of notebooks and into systems that businesses actually run on.

I stay close to the code — daily pairing with our AI engineers, architecture reviews, prompt and toolchain design, and hands-on prototyping across LangChain, PyTorch, FastAPI, and n8n. Machine learning since 2015, generative and agentic AI since 2022, software in production since 2010.

S02Checkpoints

training_run: 2010 → now · 3 checkpoints saved

ckpt_032018 — now

CTO & AI Director — MCCM Innovations

Leading the company's AI evolution end to end: machine learning and deep learning platforms from day one, enterprise GenAI programmes since 2022.

Agentic orchestration — planner–executor pipelines with tool use and supervisory loops for operations teams.
Document intelligence — extraction, classification, and validation of invoices, customs records, and legal dossiers.
RAG assistants — grounded chat over handbooks, regulation, and product docs, with human-in-the-loop feedback.
Computer vision — infrastructure inspection and manufacturing QA with custom dashboards per client.
Forecasting — energy demand, industrial production, and sales, with model governance and live inference.
GenAI for regulated teams — clinical reporting and legal review with audit trails and evaluation frameworks.

ckpt_022015 — 2018

AI Research & Software Engineer — University of Seville

Applied research across deep learning, computer vision, and ubiquitous computing.

Traffic-sign recognition combining HOG, SVM, and CNN approaches — ranked #1 on GTSRB.
Time-series forecasting and anomaly detection on live data streams.
Human-activity recognition with CNNs, RNNs, and LSTMs; e-health and e-learning Android systems.

ckpt_012010 — 2015

Full-stack web & mobile engineering

High-traffic PHP platforms, Android marketplaces, municipal social-services systems, and innovation consulting for SMEs — including co-founding Yobapp, an Alfacamp-winning mobile app builder.

Education · training data

2016 — 2018 PhD in Computer Science — Artificial Intelligence, University of Seville

2007 — 2016 BSc, MSc & University Degree in Computer Science, University of Seville

S03Research

retrieval: 3 peer-reviewed papers

[001]

Deep neural network for traffic sign recognition systems

Neural Networks · 2018 · doi:10.1016/j.neunet.2018.01.005

Spatial transformers and stochastic optimisation, reaching state-of-the-art performance on GTSRB with deployment-ready inference.

[002]

Evaluation of deep neural networks for traffic sign detection systems

Neurocomputing · 2018 · doi:10.1016/j.neucom.2018.08.009

Systematic comparison of detection architectures, trading off latency and accuracy for real-time autonomous driving.

[003]

Exploiting synergies of mobile mapping sensors and deep learning

Expert Systems with Applications · 2017 · doi:10.1016/j.eswa.2017.07.042

LiDAR, imagery, and neural networks combined to scale traffic-sign acquisition pipelines for smart mobility.

S04Capabilities

tool registry · 4 modules loaded

AI & Machine Learning: LLM orchestration, agentic workflows, RAG pipelines, computer vision (YOLO, RegNet), NLP, OCR, and MLOps for regulated industries — PyTorch, TensorFlow, LangChain.
Cloud & Platform: AWS, Azure, and Google Cloud; serverless and containerised deployments, CI/CD, observability; n8n automations and MCP server integrations; SQL and NoSQL at scale.
Product & Engineering: Python-first delivery with FastAPI, data pipelines, microservices, and agent control planes; React and Next.js interfaces when the team needs speed.
Leadership: Daily coaching and pairing with AI engineers, architecture reviews, and stakeholder alignment from kick-off to outcome review.

S05Contact

channel: open

Building something that needs to see, read, or decide?

Enterprise AI partnerships and advisory.