Over three decades of experience in practice.
A small sample of use cases are below.

2019
Specialty care hospital


2019
Law firm



2018
Primary care clinic

​
​2018
​Automotive company



2017
Primary care clinic


​
​2017
Behavioral health clinic



​2016
Hedge fund


2015
Food manufacturing


2014
Consulting firm


2014
Food manufacturing



2013
Academic hospital


​2012
Pharmaceutical company


2011
​Employee rewards


2010
Biotechnology company


2009
Private equity



​2008
Pharmaceutical company

2007
Biotechnology company

2006
​Pharmaceutical company

2005
Pharmaceutical company
​
2004
​Investment Management Company 

2003
Energy company

2002
FASB

2001
Aerospace Company
​

1999
Pharmaceutical company

1997
​Pharmaceutical company


1996
Printing press manufacturer

1995
Propane Company

1990
Private Company

1988
Heavy Equipment


​​1986
Northwestern University


1985
Indian Institute of Technology
​

​Scheduling physicians and care providers by forecasting patient flow, no-shows and other aspects.  Predicting and managing readmissions to acute care venues. We expect to reduce readmission related costs by 20%.

​Developing human resources across the firm by predicting economic outcomes of individuals and groups, selecting best personnel to optimize HR portfolio, and providing decision guidance on staffing and scheduling. We expect to improve hiring efficiencies and productivity by 10%.

Risk clustering of patients based on EMR, claims, hospital data and scheduling of high-risk patients to avoid high-cost interventions given physician availability and patient behavior. We expect to increase revenue by 10% by better management of patients needing more care.

​Forecasting of SKU demand for a manufacturer using machine learning to substantially reduce forecast errors produced by the ERP system and the use of the forecasts to reduce inventory. We demonstrated reduction in forecast errors by a factor of 5 compared to the ERP system resulting in 15% reduction in inventory.

Prediction of the onset of diseases before diagnosis and the probability of improvements, post-diagnosis for proactive interventions to prevent diseases and to improve outcomes by better care management. We expect to improve payment to physicians by 20% in hypertension.
​
Design of a system for automated matching of patients to therapists to improve outcomes, reduce the length of stay and number of visits and overall reduce the cost per episode. We demonstrated a 16% reduction in length of stay and 10% reduction in number of visits by programmatic matching.

Prediction of the probability of litigation for specific patents and the ikely number and timing of litigation based on historical patent data and litigation outcomes. We built models that can correctly predict litigation 90% of the time. 

Prediction of the size and quality of agricultural inputs coming from across the country into a manufacturing process and the optimization of manufacturing and packaging plans. We demonstrated reduction in waste of 8% by optimized selection.

​Prediction of the probability of approval of specific patent applications as well as the probability of maintenance of issued patents using unstructured data from applications and filings. We built models with prediction accuracy of over 86% in both cases.
​
​Prediction of food and workplace safety events and productivity in food manufacturing and the design of a manufacturing plan to maximize productivity and minimize safety risk. By programmatic creating of manufacturing plans, we should an enhancement of productivity by 5% and a reduction in workplace safety issues by 45%.

​​Prediction of costs savings by providing counseling to geriatric patients at intake based on patient characteristics and the selection of patients recommended for counseling.We showed a reduction of $1600/patient/episode if implemented across the portfolio.

Prediction of protocol deviation probabilities in clinical trials in R&D at the inception of studies and the design of protocols including size and locations to minimize all deviations. We showed that deviations can be predicted correctly in 75% of the instances.

Prediction of redemption patterns by employees given rewards using historical patterns, the design of the best reward features and the value of the overall portfolio due to breakage. We demonstrated humans can be substantially removed and the process can be delegated to AI.

​Predictive model for the timing and size of clinical trials, optimum design of trials considering resource constraints, availability of investigators and other constraints. The models were able to correctly guess the parameters over 75% of the time.

​Predictive and stochastic analysis of inputs in concentrating solar power plant for the valuation of the hybrid plant and the prediction of optimum operating policies to maximize value. These analyses cannot be conducted by traditional analyses as typically performed in the industry.

Prediction of the probability of success, expected duration before failure and overall resource needs and rate of spending using program characteristics and historical data.

Prediction of outcomes in various available legal strategies in a patent infringement case and the selection of the best strategy by the economic valuation of all uncertainties.

Prediction of manufacturing demand of R&D substances, optimum internal manufacturing capacity and the design of outsourcing contracts considering various features.

​Optimization of manufacturing batch sizes by predicting demand patterns from historical data and considering logistical constraints and value loss due to possible delays.

Fully automated hedge fund that collects information from the web to make trading decisions, executes trades through ECNs and dynamically learn from past trades.

Predictive analytics and economics for the pricing of underwater pumps and the design of insurance contracts using historical data.

Economic valuation of Employee Stock Options by predicting early exercise based on historical data on employee behavior and known constraints on liquidity and exercise timing.

​Economic analysis of next-generation space vehicle program to assess societal benefits, risk, and value of design choices with diverse revenue objectives including space tourism and defense.

Development of a valuation process for complex R&D projects as a basket of interacting derivatives, technical risks and cash flows using simulation and dynamic programming.

Creation of one of the largest simulation model in pharmaceutical R&D, capable of predicting outcomes of projects based on characteristics and real time statistical budgeting of $4.5 billion/year.

Optimization of the manufacturing timing, quantity, and design features by predicting demand patterns, price elasticity, and the actions likely taken by competitors.

Prediction of optimal pricing for a propane distributor and the creation of an intelligent system that can push such pricing to 600 retail locations on demand and on frequency.

​​Prediction of treatment intensity for patients to leverage physician time through risk clustering of patients and using physician extenders to improve practice economics.

Automated creation of adaptive mesh for finite element and computational fluid dynamics analyses of aircrafts, automobiles and power plants for failure analysis of equipment and components. 

Artificial Intelligence application in graduate school engineering education to accelerate design intuition of emerging students by computer interaction using the first IBM PC. This is one of the first AI programs in education.

​Intelligent agent for optimizing the design of industrial buildings given the requirements and specifications by random search of the design space using IBM 370 . This is one of the first AI programs in engineering design optimization.