AI Systems Architect · Director of Artificial Intelligence
Architect & Director of AI — 7+ years shipping AI in production.
Agentic AI · LLM infrastructure · RAG · Generative AI · Computational neuroscience foundation.
I design and ship autonomous agentic systems on top of LLMs: orchestration, persistent semantic memory, hybrid RAG over vector stores, generative image production, and cost/latency optimization. I blend rigorous scientific thinking with product-minded execution across single- and multi-agent architectures.
danigdominguez@gmail.com · Madrid, Spain
English — Full professional · Spanish — Native / bilingual
- Cajal · electrophysiology + interpretable ML
- Industrial ML · recsys · vision · finance stacks
- Multimodal generative platforms
- Agentic orchestration at scale
Research → Agents
Machine-readable résumé: /resume.json
Capabilities
Agentic AI
- Single & multi-agent architectures
- Task decomposition & agent routing
- Persistent semantic memory design
- Agent interfaces (advanced chat & non-chat copilots)
LLM infrastructure
- Orchestration stacks
- Latency / throughput / spend optimization
- Production rollout & versioning
RAG & retrieval
- Hybrid retrieval pipelines
- Vector databases (OpenSearch & similar)
- Embeddings pipelines (multimodal when needed)
Generative AI
- Image generation in production
- Multimodal classification & grounding
- Fine-tuning for in-the-wild scenarios
Classical ML / vision
- Recommendation & reinforcement-learning systems
- CNN / NLP classification & anomaly detection
- Fraud detection & synthetic-content detection pipelines
AI strategy & leadership
- Team leadership (up to 5 AI experts) — player-coach model
- AI roadmap & strategic planning
- Cross-functional stakeholder communication
- Science communication, oratory & technical writing
Experience
Director of Artificial Intelligence
BillionHands · NextChance Invest
Designing and implementing autonomous LLM-backed agent stacks for structured knowledge acquisition and automated decision pipelines.
- Agentic AI architectures (single- and multi-agent).
- Orchestration kits: task decomposition, routing, durable memory contracts.
- Hybrid RAG over vector backends with pragmatic eval harnesses.
- Interfaces that feel like assistants, copilots, and internal tooling — not gimmicks.
- Generative image models productionized alongside LLM infra.
- Latency/cost knobs for inference and routing across providers.
- Led a team of up to 5 AI experts — player-coach across research and engineering.
- Up to 50% improvement in model cost, efficiency, and output quality through systematic optimisation.
- Agentic AI
- LLM
- RAG
- GenAI
Director of Artificial Intelligence
Multimarkts
AI strategy plus hands-on platforms for multimodal search, inspiration retrieval, contextual content.
- Company-wide IA roadmap anchored to monetizable search moments.
- Multimodal embeddings (image + text) at scale.
- Operationalized OpenSearch + vector backends.
- Generative image & motion experiments moved into production pipelines.
- Reinforcement-learning inspired recommenders powering discovery.
- Led a team of up to 5 AI experts; up to 50% model improvement in cost/efficiency/quality.
- Multimodal
- OpenSearch
- RecSys
- GenAI
Lead Data Scientist
NoFakes
Fraud, synthetic-media, abusive-content detection backed by NLP + behavioral signals.
- Semantic NLP classifiers tuned for nuanced policy targets.
- Behavioural anomalies for detecting coordinated bot activity.
- Synthetic-text attribution models (GPT-like generators).
- End-to-end detection/classification pipelines for fraud ops.
- NLP
- Safety
- Fraud
Technical Co-founder
Adsviu · NextChance Invest
Conceived and built the computer-vision MVP that originated the company — cookieless contextual advertising with proprietary AI.
- End-to-end MVP development (computer vision + product contextualisation) — 0-to-1 validation.
- Technology that enabled ×3 CTR, ×4 Brand Awareness for fashion and editorial clients.
- Clients included a leading editorial group (14 outlets, +15M unique visitors).
- Computer Vision
- AdTech
- 0-to-1
- Founding
Senior Data Scientist
BillionHands AI Fashion
Vision + NLP product models for garment understanding in the wild.
- CNN + NLP garment recognition architectures.
- Multimodal classification (visual + textual metadata).
- Bayesian hyperparameter optimisation for latency-sensitive ensembles.
- Product reached +500,000 downloads on Android (visual fashion search / shop-the-look).
- Computer Vision
- Multimodal
Data Scientist
Dive.tech
Consulting-grade delivery of predictive & vision workloads.
- Personalised recommender stacks (sequence + topic hybrids).
- Industrial QA vision for defect spotting.
- Person-identification workloads with privacy guardrails.
- Streaming video analytics pipelines on AWS.
- RecSys
- Vision
- AWS
Professor of Artificial Intelligence
Afi Escuela de Finanzas
- Teaching CNN + RNN blocks inside the formative AI syllabus.
- Teaching
Data Scientist · Financial Client
Novabase
- Reinvestment RL strategies for discretionary desks.
- NLP on scanned contractual packages.
- Signature authenticity via CNN ensembles.
- Finance
- RL
- CNN
Science communication & broadcasting
Freelance / portfolio
- >10k YouTube subscribers, >100k views on educational neurochemistry/neurophysics content.
- Weekly columns via national publishers (La Razón & peers).
- Science communication spanning blogs, national press, YouTube, and editorial books.
- SciComm
Predoctoral researcher · Computational neuroscience
Instituto Cajal · CSIC
- Brain rhythmopathies modelling + EEG / field-potential pipelines.
- Large-scale spike sorting + preprocessing at scale.
- Interpretable ML for electrophysiology datasets.
- ~17 publications, ~900 citations, including Nature Neuroscience (2015, ~278 citations).
- Research
- ML
Stage magician & lecturer
Independent
- Theatre illusions plus large-format kids shows sharpened narration and stage presence.
- Oratory
Education
M.Sc. Biofísica · Electrofisiología
Universidad Autónoma de Madrid · 2011–2012
B.Sc. Bioquímica
Universidad Autónoma de Madrid · 2009–2011
Diplomatura · Química
Universidad Autónoma de Madrid · 2006–2009
Certifications
DeepLearning.ai / Coursera
- Structuring Machine Learning Projects
- Neural Networks & Deep Learning
- Improving Deep Neural Networks
- Sequence Models
- Deep Learning Essentials
Publications & speaking
- Matemáticas y Neurociencia
- Neurociencia · Estructura y funciones del cerebro
SciComm spanning national press columns, long-form neuroscience explainers, and video essays.
~17 publications, ~900 citations — computational neuroscience at Instituto Cajal (CSIC).
- Determinants of different deep and superficial CA1 pyramidal cell dynamics during sharp-wave ripples — Nature Neuroscience (2015, ~278 citations)
- Distinct Mechanisms of Over-Representation of Landmarks and Rewards in the Hippocampus — Cell Reports (2020)
- KCC2 Regulates Neuronal Excitability and Hippocampal Activity via Interaction with Task-3 Channels — Cell Reports (2019)
- Altered Oscillatory Dynamics of CA1 Parvalbumin Basket Cells during Theta-Gamma Rhythmopathies — eNeuro (2016)
Beyond work
Earlier career as a touring stage magician honed storytelling, improvisation, and audience calibration — competencies that map cleanly to stakeholder comms today.