John Kalantari, Ph.D.

AI/ML Scientist — Center for Individualized Medicine, Mayo Clinic

Research Interests

I work on:
Computational Systems Biology
Artificial Intelligence
Machine Learning
Complex Adaptive Systems
Algebraic Topology
Data analysis tools
Bayesian Statistics
Numerical simulations
Visualization
What I like to do: Analyzing data from all sorts of sources, and creating tools to support others' ability to do the same; Creating tools designed to enable the design of data-rich systems in novel interdisclipinary research situations. My focus and the focus that I encourage in others, is to use algorithms that can be clearly communicated and interpreted in the organizations that deploy them.

Current Projects


Pop-Up Restaurant for Inverse Reinforcement Learning Probabilistic Machine Learning for Modeling Complex Causal Mechanisms

Past Projects


Simplicial Grammar Computational data abstraction for modeling multi-scale systems
Stochastic Grammars
Java
Algebraic Topology
Combinatorics
General-Purpose Modeling
Beta

Simplicial Grammar (SG) is a novel computational abstraction that extends traditional grammars used in automata theory and linguistics. An SG is a Bayesian nonparametric modeling formalism that describes complex pattern classes and the process of their formation as a topological computation. This formalism is used to describe the explicit, hierarchical, and multi-dimensional structure of patterns inherent in sequence data.

GitHub

SYNACX Syntactic Analysis of Complex Systems
Artificial General Intelligence Framework
Unsupervised Sequence Learning Algorithm
Hierarchical Generative Modeling
Bayesian Nonparametrics
Beta

SYNACX is an unsupervised learning algorithm that extracts high-level, complex abstractions from time-series data through a hierarchical learning process.

GitHub

Cellular Event Prediction (CEP) Temporal analysis of cellular morphology
Cell Biology
Predictive Analytics
Java
In-silico modeling
Beta

CEP is a software package for modeling and analyzing cellular morphology data. It consists of a fast, parallelized Java library (which you can easily integrate into your own project), an R package providing a high-level interface to the library, and an easy-to-use web interface for interactive analysis and plotting.

CEP offers incremental modeling and visualization, Markov-Chain Monte Carlo and uncertainty estimation, cross-validation.

GitHub