Liner Partners With KAIST di-Lab to Pilot AI Agent Use in Policy Research
The AMW Read
Extends Liner's academic research agent into a new pilot vertical via a university MOU, incremental for a known player with no funding or scale evidence.
Liner Partners With KAIST di-Lab to Pilot AI Agent Use in Policy Research
Liner, the AI agent startup led by CEO Jinwoo Kim, signed a memorandum of understanding with KAIST di-Lab, the Digital Innovation and International Development Lab run by Professor Kyung-Ryul Park at KAIST's Graduate School of Science and Technology Policy. The MOU, signed September 21 at KAIST's Center for Science and Technology and Global Development, applies Liner's academic research agent, Liner Scholar, to international-development and science-and-technology policy research. The plan has three phases: a pilot where di-Lab researchers use Liner Scholar across data collection, policy analysis, and paper writing, with both sides jointly analyzing how it affects research productivity and cross-country/region gaps in research capacity and AI access; a domain knowledge graph plus a verification agent that checks whether structured policy information traces back to its original sources; and, longer term, a policy-simulation environment modeling stakeholders — donor institutions, recipient governments, field implementers — as AI agents to explore likely effects of a given policy or program.
The deal extends Liner Scholar beyond its earlier materials-science research use case into a second vertical, and lands weeks after Liner's Series C and its NPU-infrastructure partnership with Rebellions — per the AI Market Watch index, Liner has raised $36.4M to date across a roughly 5,000-company coverage set, not a census. Together the three moves read as an academic-agent vendor testing whether its core retrieve-verify-synthesize loop generalizes past paper search into evidence-grounded policy analysis and, eventually, agent-simulated stakeholder behavior.
For builders, the reusable piece is the verification agent binding generated claims to primary sources — the harder half of a research-agent product. For investors, this is a university pilot with productivity data as the deliverable, not a revenue signal, so the read-through depends on whether di-Lab's findings convert into paying government or institutional contracts.
