AI & Big Data Lab · Asian Development Bank · 2026

AI at the Asian Development Bank

Applying AI to development challenges across Asia and the Pacific

Jude Teves
Lead Data Scientist & AI Engineer
Asian Development Bank
linkedin.com/in/judeteves

1 / 13

Who's talking

Jude Teves

Lead Data Scientist & AI Engineer · Asian Development Bank, AI & Big Data Lab

Computer Science and Engineering background; career spanning software engineering, startups, data science, ML engineering, data engineering, and academia — now focused on AI engineering and research

Industry experience across tech companies, financial institutions, government agencies, multilateral organizations, and universities

Core focus on model development and operationalization — from research to production — alongside work on responsible AI and institutional alignment

Sports junkie, traveler, educator and community mentor, geek, and generalist

2 / 13

69

member countries,
50 from Asia-Pacific

$29B+

in commitments
in 2025

60

years of development
financing since 1966

At this scale, every AI system we work on touches real decisions — infrastructure investments, climate adaptation strategies, sovereign lending. That changes what responsible AI actually means.

3 / 13

Who we are

The AI & Big Data Lab — how we work

⚖️

Ethics

Responsible AI

Defining the values, principles, and governance structures that determine how AI is built and used across ADB.

🔬

Science

AI Research

Advancing what's possible — model evaluation, knowledge representation, and improving the intelligence behind our systems.

⚙️

Engineering

AI Infrastructure

Operationalizing research into production — sovereign AI systems, data pipelines, and infrastructure that ADB owns and controls.

The most interesting work happens where these three areas overlap — and that's where today's projects come from.

4 / 13

The lab's work

Some of our projects

1

Climate & Risk

Disaster Risk Analytics Explorer

Geospatial AI platform for risk-informed decision-making across Asia and the Pacific.

2

Knowledge & AI

Genie

ADB's enterprise generative AI platform — institutional knowledge, accessible to all staff.

3

Credit & Finance

Credit AI

AI-powered platform for credit risk assessment and workflow augmentation — from screening to decision support.

4

AI Governance

Responsible AI Framework

Governance tools and evaluation systems that underpin everything else the lab works on.

5 / 13

Disaster Risk Analytics Explorer

From climate data to decision-ready insights

Case · Tonga Pilot

Pre-collected climate and disaster risk data combined with AI-powered tools — giving Tonga's planners access to hazard maps, financial impact estimates, and resilience scenarios for the first time in an interactive format.

Scenario planning across 2035, 2050, and 2070 climate timelines

Five hazard types — flood, sea level rise, wind, tsunami, earthquake — in one interface

AI-generated insights that translate technical risk data into investment decisions

Sector research, guidelines, and infrastructure reports layered onto the map

The design principle: the countries most exposed to climate risk shouldn't need months of expert consultation to understand it. Make the analysis accessible to the people who need to act on it.

Why it matters

Pacific Island nations bear the highest climate risk but have the least capacity to analyze it.

This platform doesn't just provide data — it bridges the gap between complex climate science and the operational decisions governments need to make now. The Explorer makes hazard analysis interactive, visual, and accessible — without requiring specialized GIS or climate modeling expertise.

6 / 13

Disaster Risk Analytics Explorer · resilienceexplorer.adb.org

7 / 13

Genie · genie.adb.org

Democratizing institutional knowledge

📄

Research & Analysis

Staff can ask questions across ADB's archive — precedents, project outcomes, sector reviews — and get answers grounded in actual ADB documents, not the open web.

✍️

Content & Drafting

Policy memos, project summaries, and reports drafted with context from ADB's institutional knowledge base — not just a generic LLM.

🔗

Policy & Compliance

Retrieves and summarizes relevant policy content from connected knowledge sources, helping users understand requirements without navigating dense documentation.

The design principle: institutional knowledge shouldn't depend on how long you've been here. If you know the right question, you should be able to get an answer — grounded in ADB's own documents, not the open web.

8 / 13

Genie · genie.adb.org

9 / 13

Credit AI

An End-to-End AI Credit Platform

Credit analysis in development finance is complex and high-stakes. Every transaction involves layered financial data, evolving sector risks, and significant capital decisions — all under time pressure, with real consequences for real projects.

The platform is built to augment the credit workflow end-to-end — from data ingestion and automated screening through document generation, risk analytics, and decision support. Every output is explainable and auditable.

The design principle: AI handles the preparation work. The analyst owns the judgment.

Faster workflows Consistent decisions Earlier risk signals

Core Components

AI Credit Engine

Automated screening, risk identification, and red flag detection across transactions

Data Integration

Single source of truth combining internal and external data with automated ingestion

Document Automation

AI-assisted generation of credit memos, summaries, and structured analyses

Decision Support

Financial summaries, indicative risk ratings, sector and country insights

Collaboration Layer

Shared platform bridging deal teams and credit officers, reducing information gaps

Learning System

Human-in-the-loop AI continuously refined by senior credit officers

10 / 13

Responsible AI

Making AI Trustworthy by Design

AI deployed at the scale of a multilateral development bank affects real decisions in real countries. The goal of ADB's Responsible AI initiative: ensure every system built here is not just capable — but trustworthy, explainable, and aligned with ADB's mission.

That intent is operationalized through 8 principles applied to every AI initiative the lab works on — from design through deployment.

Responsible AI isn't a separate workstream. It's the foundation every project in this talk is built on.

ADB's Responsible AI Framework — 8 Pillars

🌍 People & Planet First

Avoid harm; anticipate and mitigate impacts on people and environment

🛡️ Safety & Reliability

Within safety parameters; consistent performance to standards

🔍 Explainability

Explain reasoning at appropriate level; document limitations

🔒 Privacy & Security

Data privacy by design; safeguards against threats and unauthorized access

⚖️ Fairness

Minimize bias; inclusive design; testing across the full lifecycle

👁️ Transparency

Disclose data, model, decisions; identify AI-generated outputs

📊 Data Integrity

Quality, governance, lineage — reliable and fit-for-purpose data

👤 Accountability

Human oversight across AI lifecycle; clear roles and decision records

11 / 13

A thought to leave you with

At this scale, how you build
matters as much as what you build.

ADB's work operates at the scale of nations — climate systems, credit decisions, knowledge infrastructure built for dozens of countries and hundreds of millions of people. That scale raises the stakes on every design choice.

Responsible AI isn't a compliance layer added at the end. It's what makes it possible to deploy at scale, maintain trust, and sustain impact over time — and it has to be baked in from the start.

The opportunity here isn't just to build impressive AI. It's to show that AI built carefully can reach further, last longer, and do more good than AI built fast.

12 / 13

Ask me
anything.

linkedin.com/in/judeteves

13 / 13