avianna
We do applied research at the intersection of AI systems, agents, and the future of software work.
Advisory
Avianna advises a small number of startups and enterprise teams on AI strategy, agent workflows, and architecture, grounded in the lab’s research. For advisory or collaboration inquiries, contact the lab.
Research areas
AI is becoming systems that plan, decide, and act. Three questions decide whether agents work in the enterprise:
Economics
What an agent costs and returns, and why so many pilots never pay back.
Reliability
When agent work can be trusted to run, how it fails, and how it recovers.
Alignment and governance
Keeping agents inside the organization’s intent: authorized, overseen, reversible.
LeetMath
A lab prototype from our mathematical-reasoning work: elite math training for adults with real-world intuition and limited time. The first track is probability — forty problems from sample spaces to the probabilistic method, each tagged by the kind of hard it tests, with hint ladders, wrong-answer diagnosis, and spaced review. Start the probability track →
About
Avianna is an independent research lab founded by Tushar Madan and Rishubh Khurana. The lab studies how AI systems, agents, software, and organizations work together safely, reliably, and intelligently, and advises a small number of startups and enterprise teams on AI strategy. Avianna builds Concord (contracts for agent actions), Lattice (the worker layer of the governance stack), and LeetMath (probability training for adults). Avianna is not affiliated with the drone-software company previously associated with this domain name.