Malhar Chaudhari

Granted patent describing methods to identify causal relationships between time series in hierarchical node structures, generate models to estimate causal coefficients, and produce ranked outputs of nodes that influence target nodes.

Patent application describing methods for generating natural language outputs for model simulations of hierarchical time-series node networks, including templates, semantic tag population, and integration with language models.

Research on using rateless fountain coding strategies to mitigate stragglers in distributed matrix-vector multiplication; shows asymptotically near-ideal load balancing with minimal redundant computation and reports experimental speed-ups over uncoded schemes.