At Kimi AI, I author image-based biology reasoning tasks where the answer must come from interpreting a figure alongside its scientific context. I check for alternative valid interpretations and write expert reference answers that document each reasoning step.
At Handshake AI, I turn primary research papers and supplementary data into single-answer biology problems designed to test genuine reasoning. I target challenges such as multi-hop inference and constraint interactions, rather than obscure fact recall.
At Turing, I contributed to frontier-lab evaluation programmes, including NVIDIA SciCode task generation and Claude Opus 4.8 evaluation. I decomposed research problems into scored tasks and golden solutions, and built bioinformatics problems on real datasets with deterministic answers.
Before moving into AI training and evaluation, I was a Senior Research Fellow at Gujarat Biotechnology Research Centre, where I worked across metagenomics, whole-genome sequencing and AMR genomics. I led research from sequencing data to a commercial probiotic technology transfer, and my research includes six peer-reviewed publications.
