Research & reports

Benchmarks, blueprints, and evaluation research developed across public systems in 28 countries.

August 2026

The 2026 Global Sovereign AI Benchmark: Institutional Readiness Across 28 Nations

An empirical investigation into national compute self-sufficiency, public dataset stewardship, and algorithmic governance across low- and middle-income economies.

Authors: Dr. Francis Xavier Rathinam, Nakul Jain, Deepa Karthykeyan, Dr. Sudhanshu Joshi

Read Executive Abstract
Expand
This benchmark provides the first comprehensive empirical evaluation of national sovereign AI readiness. Analyzing institutional infrastructure, data residency mandates, and operational AI deployments across 28 governments in Africa, Asia, and Latin America, the study reveals critical vulnerabilities in public-sector procurement while outlining proven architectural remedies for long-term algorithmic independence.

Publications

July 2026

Algorithmic Fairness in Primary Education: A Multi-Dialect Evaluation Protocol

Standardized methodology for evaluating speech-to-text literacy models across non-standard dialects, acoustic variations, and gender representations in public classrooms.

Authors: Dr. Harsh Vats, Dr. Francis Xavier Rathinam, Anupama Ramaswamy

Read Executive Abstract
Expand
This protocol provides public education authorities with an empirical framework for evaluating speech-enabled literacy assessments. We detail quantitative testing procedures across ~8,000 recordings in Hindi and Gujarati, highlighting methods to detect and remediate dialectal bias prior to high-stakes classroom adoption.

May 2026

Municipal Utility AI Sandboxes: Framework for Empirical Pre-Procurement

Operational blueprint for establishing secure, data-isolated environments where public utilities can benchmark commercial leakage detection and pressure models before contracting.

Authors: Kowshik Ganesh, Suvabrata Roy, Vijay Bhalaki

Read Executive Abstract
Expand
Procuring commercial AI for municipal water utilities carries severe financial and operational risk when claims are tested only post-contract. This framework outlines the architecture, data de-identification, and validation metrics needed to run a 90-day sandbox using real SCADA telemetry.

December 2025

Farmer-Centric Data Governance: Legal, Technical, and Consent Architectures

Architectural specifications for protecting smallholder data sovereignty, establishing collective consent protocols, and preventing extractive data practices across agrifood ecosystems.

Authors: Dr. Rajesh Khanna, Suvabrata Roy, Dr. Sudhanshu Joshi

Read Executive Abstract
Expand
Smallholder farmers frequently generate valuable telemetry without retaining ownership or receiving fair economic value. Developed in collaboration with multilateral funders, this whitepaper defines legal consent contracts, API access tiers, and tokenized governance primitives that enforce farmer rights.

October 2025

MSME AI Readiness Diagnostic: Survey of 1,200 Micro-Manufacturers

Field findings on equipment telemetry maturity, edge compute constraints, and working capital barriers to automated visual quality control in precision manufacturing.

Authors: Vijay Bhalaki, Nakul Jain, Ankit Chatri

Read Executive Abstract
Expand
Drawing on on-site diagnostics across 1,200 manufacturing enterprises in 6 industrial clusters, this brief details the actual technical and financial readiness for AI deployment. We outline practical modular upgrades that yield immediate ROI without capital-intensive machinery replacement.

Looking for specific benchmark data or research?

Our team conducts independent evaluations and diagnostics for public programs and development partners.

Contact research team →
Citation copied to clipboard