My work centers on one problem: when one model judges another’s output, can that judgment be trusted, where does it fail, and what fixes it? I study it in research and build against it in production as a data scientist and AI engineer.
Currently a Data Scientist / AI Engineer at Climate Clean Solutions & LowPropTax, where I own production for a multi-service agent platform on AWS. Previously at Fusemachines and NAAMII. B.Sc. in Computer Science and Information Technology, Tribhuvan University.
Research interests: evaluation reliability, LLM-as-judge, error propagation in multi-agent systems, AI safety and security.
Publications
Neither Blinding Nor a Jury, Neither Capability Nor Strategy: What Brings a Verifier to Reliability
Which Verifier Catches What? Mapping Error Severity to Verifier Strategy in Multi-Agent LLM Pipelines
Recent
- Sep 2026 · Paper accepted at the NeurIPS 2026 Workshop on Trust-AI-Eval; extended version submitted to ICLR 2027.
- Aug 2026 · Fault-injection study selected as a top AI/ML project at SISTER 2026.
- Jan 2026 · Joined Climate Clean Solutions & LowPropTax.