From raw chaos to decision-grade AI.

AI and Data Science Projects

• Climate Evidence Engine 2026

  An AI research pipeline for the University of Zurich that reads sustainability reports, annual reports and SEC 10-K filings and turns them into structured, source-linked evidence.

  5,880 evidence rows · 392 STOXX 600 firms · 15 climate indicators

  • LLM-as-judge
  • ClimateBERT
  • Python
  • Reproducibility

  Full technical write-up →
• AI Early Warning Systems 2025

  Deep-learning and ML models that forecast tropical storms, droughts and floods for the UN Office for Disaster Risk Reduction, feeding a live investment observatory for disaster-risk financing. The work won an international award and was presented at GP2025 in Geneva.

  150+ countries in the governance taxonomy · 4 global data sources: GIRI, EM-DAT, World Bank, OECD CRS

  • PyTorch
  • XGBoost
  • Time series
  • Power BI

  Watch the GP2025 presentation ↗ Virtual exhibition ↗ Code on GitHub ↗ Full technical write-up →
• Mathimatikos.xyz 2025

  An AI-native STEM reasoning engine. Multimodal OCR and RAG pipelines break a problem down into step-by-step, validated explanations with rendered maths and graphs, instead of one opaque answer. Queue-backed services, typed API contracts and usage metering keep it running in production.

  5 subjects: maths, statistics, finance, physics, CS · 3 input modes: typed, photo, handwriting

  • Multimodal OCR
  • RAG
  • TypeScript
  • PostgreSQL
  • Redis queues

  Visit mathimatikos.xyz ↗ Full technical write-up →
• BetterPrompt 2025

  A Chrome extension and SaaS backend that rewrites rough input into expert-level prompts: intent analysis, academic/technical/creative classification, tone control and guided refinement, inserted straight into the tool you're already in. The backend routes between Gemini and Anthropic behind a secure proxy, with quota metering and subscriptions.

  8 AI tools supported · 4 prompt classes · secure key proxy, no keys in the browser

  • Chrome MV3
  • Node.js
  • LLM orchestration
  • PostgreSQL
  • Payments

  Get it on the Chrome Web Store ↗ Full technical write-up →
• Humanitarian Dashboards 2024

  Production food-security, displacement and shock-monitoring pipelines for UN Women, feeding interactive maps and comparison tools that rank where support is needed first. Includes a documented Shock Exposure Index: component scoring, weighting, missing-data rules and interpretation limits.

  9 public data sources: WFP/HDX, IPC, WDI, FAOSTAT, UNHCR, OCHA FTS, ILOSTAT, IDMC/IOM, Findex

  • Python
  • MapLibre
  • Chart.js
  • Data governance

  Full technical write-up →
• Power BI Case Studies 2025

  Independent Power BI work: three synthetic-data demonstrations styled after the kind of organisation each one imagines serving, plus one built on real, anonymized client data.

  4 live reports · audit compliance, scientific monitoring, talent analytics and maritime operations

  • UNICEF-style audit compliance ↗
  • IAEA-style water science ↗
  • IFC-style talent analytics ↗
  • Signal Maritime cargo analytics ↗

  All dashboards on GitHub ↗ Full technical write-up →

The Journey

1. 2026University of ZurichAI Engineer & Research Associate
2. 2023 – 26United NationsUN Women · UNDRR · IOM
3. 2021 – 23BI Solutions GroupAdvanced Analytics & AI Consulting
4. 2020 – 21IAEA, ViennaInformation Systems Analyst
5. 2019 – 20Hellenic Institute of Cultural DiplomacyStatistician & Data Analyst

Worked with

• UN Women
• UNDRR
• IOM
• IAEA
• UZH
• Fujitsu
• LG
• Nespresso
• Collins Aerospace
• Coca-Cola
• PepsiCo

Talks & Writing

• Jun 2025 · Geneva Global Platform for Disaster Risk Reduction (GP2025) Award-winning AI early-warning models, presented at the UN General Assembly-mandated forum and the Early Warnings for All multi-stakeholder forum. Watch ↗
• Dec 2024 · Athens 22nd Investment & Stock Market Conference Guest speaker, Session 5: investing in new technologies and AI, and crypto-assets under the new MiCA regulation. Watch ↗

Data literacy articles

• How the Charlotte Hornets used big data to score big with fans
• The tale of two APIs: OData vs SOAP for SuccessFactors → Power BI
• Power BI with GA4 and UA: challenges and fixes from IOM
• Faster queries and efficient data modelling in Power BI
• Harnessing OCR in data science
• A useful “igloo” tool for working with uncertainty

Teaching & consulting

I tutor and train in statistics, econometrics, Power BI, Python and SQL, for students from LSE to Queen's Smith School of Business, and consult in data science, mathematics and finance.

Engineering rigour rooted in pure mathematics.

Mathematician by foundation, system architect by practice. I design AI systems that withstand real-world scrutiny: deterministic evaluations, verifiable data pipelines, and decision interfaces that non-technical leaders can act on. Read my M.Sc. thesis ↗ · Meet the mascot →

GenevaPresented AI early-warning models at GP2025, the UN Global Platform for Disaster Risk Reduction.

9.3 / 10M.Sc. in Operational Research, Analytics & Decision Making. Thesis on AI and multi-criteria decision analysis.

StockholmExchange semester in Financial Mathematics during my B.Sc. in Mathematics, University of Athens.

ΕΛ · ENGreek native, fluent in English. Working remotely with teams across Europe.

“Ioannis is easily the best tutor/trainer I have had in my professional life.”

Gautham · Power BI student

“He's incredibly efficient, respects our company's data privacy, and consistently delivers fast, high-quality results.”

Theodore Anton · consulting client

“He taught Statistics for my LSE 2nd year course. Will hire again.”

Shalini · LSE student

• Python
• SQL
• TypeScript
• PyTorch
• LLMs & RAG
• Databricks / Spark
• dbt
• Power BI
• AWS · Azure · GCP