ALFONSO GUTIÉRREZ DE TERÁN · AI PRODUCT · BARCELONA
AI products, built from inside the factory.
Nine years managing projects in rolling stock, planning how engineering actually gets built. That is the foundation I build AI products on: systems designed to survive contact with the shop floor.
Currently: shipping two products of my own, running AI systems in production at an industrial plant, and finishing a Master's on prescriptive decision systems.
SIDE PROJECTS
Two products, both live
Things I co-founded or created, with users, data and their own consequences.
A knowledge graph of how scientific ideas actually traveled.
History of science is usually told as a parade of lone geniuses. Wikinventia maps it as a system: discoveries as nodes, influence as typed, evidence-graded edges — who read whom, what was translated, where credit was silently dropped. Built to correct historiographic bias, not just visualize it.
✓1000+ curated entries and documented relations, each tagged with its evidence type.
Co-founded and operating since October 2024. Pinwin lets people bet on real sports with settlement handled by smart contracts instead of a bookmaker's back office. It exposes the whole flow to AI agents via API, so an agent can place and claim bets the same way a person does.
✓$3M+ wagered on-chain to date.
AZURO PROTOCOL
SOLANA
POLYGON
API + MCP AGENTS
APPLIED AI
AI inside an industrial company
Systems built and deployed where the constraints are real: legacy data, physical assets, and people whose day depends on the answer being right.
IN PRODUCTION
RAG
FUNCTION CALLING
CURATED DOCUMENTATION + SYSTEM PROMPT
M365 COPILOT
Planning Compass
A retrieval-and-action copilot for enterprise planners.
Project planners lose hours hunting for the current truth across schedules, mail, documents, and enterprise procedures. This assistant answers grounded questions over the user environment and executes routine tasks through function calling on curated enterprise guidelines.
✓In production across all Alstom divisions since December 2025.
RAG
FUNCTION CALLING
CURATED DOCUMENTATION + SYSTEM PROMPT
M365 COPILOT
IN PRODUCTION
LIDAR
DIGITAL TWIN
RANKING ALGORITHM
ENDPOINT DEPLOYMENT
LiDAR × Digital Twin
Scan an assembly, get the exact part number.
Matches LiDAR scans of a physical train against its engineering bill of materials, so a defect found on the shop floor is attributed to the precise component procured. Bridging the gap between the physical asset and its material code without manual lookup.
✓API deployed matching engineering BOM data
✓submitted to Alstom Innovation Awards 2027.
LIDAR
DIGITAL TWIN
RANKING ALGORITHM
ENDPOINT DEPLOYMENT
COMPLETING 2026
SAP INTEGRATION
COMPLEXITY MODEL
PREDICTIVE ANALYTICS
PRESCRIPTIVE ANALYTICS
Predictive and Prescriptive Planning
Predicting delivery outcomes, then prescribing the optimal production portfolio.
A data pipeline pulls SAP's historical record of every rolling stock product entry and exit across the production line, scored against completeness relative to each car's last configuration issued by engineering and weighted by a project complexity model. That training set drives predictive analysis for ongoing projects and open tenders, then a Monte Carlo simulation runs across the full portfolio of parallel-line predictions to surface the scheduling scenario that optimizes cash flow and on-time delivery.
✓70% predictive accuracy today, targeting 85% in 2026.
SAP INTEGRATION
COMPLEXITY MODEL
PREDICTIVE ANALYTICS
PRESCRIPTIVE ANALYTICS
BACKGROUND
Academic background
Ten years of formal engineering education, extended into business analytics and applied AI.
2011
BSc Mechanical Engineering
Universidad Politécnica de Valencia.
Foundation in mechanics, thermodynamics and industrial systems — the basis.
2013
MEM Industrial Management
Universidad Politécnica de Valencia / Helsinki Metropolia University of Applied Sciences.
Postgraduate specialization in industrial and organizational engineering, building on the mechanical engineering base.
COMPLETING 2026
MSc Strategic Business Analytics
EAE Business School, Barcelona.
Focused on predictive modeling, experimentation, prescriptive decision systems and the governance of AI projects — the managerial layer most AI initiatives fail at.
PLANNED
PhD — prediction & prescription for industrial planning
The research question behind everything above.
How do you design the layer between a forecast and a planner's decision so that the system's recommendation is trusted, auditable and actually followed? Industrial planning is the testbed; the answer generalizes.
