Operations Research

SAMEER D. MOKASHI

Mayur Vihar Phase I, Delhi Tel: +91-11-227***** Email: ******.*******@*****.***

Career Goal: Providing pragmatic solutions to large-scale problems of industrial relevance through business analysis and math modelling

RESEARCH & WORK INTERESTS

• Work: Business analysis, Optimization models, Decision support for supply chains

• Research: Combinatorial optimization, graph theory, Mixed integer programming models, random search techniques for scheduling and routing, portfolio optimization

EDUCATION

• Ph.D., Department of Process Integration [Sep. 1995 – Nov. 1999]

University of Manchester Institute of Science and Technology (UMIST), UK

Topic: Contextual Optimization for scheduling and planning of logistics systems

• Master of Technology (5 year integrated), Department of Chemical Engineering [Aug. 89 – Jul. 94]

IIT - Bombay, India (GPA 8.3/10)

• IIT JEE All India Rank: 811

CAREER - current

• Modelling and Optimization Consultant for SCA Technologies, LLC (outsourced mode, refer Work projects – pg3) [Oct 01 – current]

 Developed a mixed integer linear programming (MIP) model for reorganization of mfg. facilities of Eagle Picher – wrote a reformulation so that part of the model has a min-cost flow type structure

 Developed a MIP model for optimizing parts routing and outsourcing decisions at Kennametal

 Maintained the McDonalds distribution and poultry optimization models

 Developed a generic framework for production models that would capture a wide variety of features relevant to manufacturing

• Developing optimization tools and models for the following domains (personal business initiative, refer Research projects - pg4) [Mar 05 – current]

 Operations scheduling

 Optimization solvers

 Investment modeling

CAREER - previous

• Senior Operations Research Analyst, SCA Technologies LLC, Pittsburgh PA, USA [Feb. 2000 – Jul 2001]

 Demonstrated savings of $30 million for a supply chain with annual business of $650 million by realigning operations and optimising buy/ sell decisions (see highlights 1)

 Directly interacted with clients like McDonalds to facilitate decision making

 Managed a modelling team (2 masters graduates in IE/OR from Michigan and Purdue) and provided functional specs to the development team

• Research interaction – Decision Science Group, Unilever, UK [Dec. 98 – Jul. 99]

• Technology transfer – Mitsubishi Chemical Corporation (MCC), Japan [Jul. 96 – Jun 97]

 Implemented an algorithm for delivery scheduling on MCCs in-house planning system (highlights 2)

 Developed a vehicle routing algorithm for MCCs polyolefin deliveries

• Analyst – Reliance Petroleum Ltd. Mumbai, India [Jul. 94 – Sep 95]

o Worked on developing a user interface for demonstrating results of refinery optimization

SOFTWARE SKILLS: C++, GAMS, ILOG–CPLEX, LINGO

PROFESSIONAL ACTIVITIES/ INTERACTIONS

• Visiting Faculty for the Industrial Engineering programme at the Department of Mechanical Engg. of IIT-Delhi [Aug 08 – current]. Conducted the following courses for masters students

o Operations Research (LP, IP, Branch and Bound, Probability Theory, Queuing, Bayes Decision Theory)

o Supply Chain Management (Strategic fit, Network design, inventory management, Distribution networks, Demand forecasting and Revenue management, Stochastic modeling of demand and supply uncertainty)

• Visiting Faculty at the IEOR inter-disciplinary programme of IIT-Bombay [Jan 03 – Apr 03]

o Conducted a course in Applied Integer Programming (LP, IP models, TSP, VRP, Branch and bound, Dynamic programming, Column generation, Dantzig-Wolfe decomposition, model tightening reformulations)

• Worked on collaborative research with industry while in academia

 Mitsubishi Chemical Corporation, Mizushima plant, Japan

 Unilever, UK

HIGHLIGHTS

1. Part of the team at SCA that won the prestigious SCM Technology Excellence Award, 2003 given by the Supply Chain Council (SCOR) to the organization that develops a methodology or product that enables superior performance in supply chain operations

2. Demonstrated savings of 30 million Yen per annum per manufacturing site of Mitsubishi, provided a basis to the company for negotiating a deal with the logistics company, reducing cost by more than 5%

3. Successful implementation of my algorithms helped in retaining Mitsubishi’s sponsorship to the Process Integration Research Consortium (PIRC) at UMIST, UK

WORK PROJECTS

 Optimized the delivery (of PVC) operations of Mitsubishi Chemical Corporation’s Mizushima plant in Japan. Developed a graph representation based heuristic to minimize the cost of delivering from their production site to various locations using lorries, trailors and railway wagons (TL deliveries). Used a min-cost flow type network model for customer orders and their precedence and designed a sequential algorithm to work on this representation

 Developed another graph based representation for vehicle routing and implemented a heuristic algorithm to work on this representation. This model was used for Mitsubishi’s vehicle routing from its depot in Mizushima to several locations in Japan. The model involved translating the routing problem to node packing and solving the node packing problem using a heuristic (developed as part of my PhD).

 Reduced the cost of producing chicken products like Nuggets, Grilled and Crispy sandwiches for McDonalds by optimizing their first processing investment decisions, line allocation to various products, synchronization of first and further processing product mix at various plants of their poultry vendors – Tyson Chicken and Keystone Foods. Developed a deterministic mixed integer linear programming (MIP) model to model McDonalds’ chicken business. Followed this up with a stochastic programming model with scenarios capturing uncertainty in the demand of 6 products and the prices of raw meat. The demand and price uncertainty was modelled by using user provided probability numbers and a limited combinatorial scenario explosion after accounting for phenomena such as cannibalization.

 Developed a MIP (mixed integer programming) model for McDonalds’ nationwide distribution system capturing outbound transportation, inventory, order processing, shipment receiving and inbound transportation costs. Used a piecewise linear approximation for the non-linear expression for safety stock. Assumed that demand is uncorrelated for different stores for the safety stock calculation.

 Helped Eagle Picher (an auto parts manufacturer) with a model to support their resource movement decisions from their US plants to new locations in Mexico. This was a multi period MIP model with a min-cost flow type reformulation capturing various production costs and one time investment/cost of work force severance, line movement, overtime expenses to make up for downtime and constraints such as those on availability of skilled engineers required for the move, budget allocated to meet one time expenses for the move.

 Optimized product routing on resources for Kennametal, a drill bits producer, at its current plants in the US. Implemented a mixed integer linear programming model to analyze its outsourcing decisions by considering trade-offs of labor cost vs. freight and constraints on notions of quality of high precision drill bits.

 Tested a demand forecasting, pricing and promotions model. The model used a combination of linear regression (log-linear with least squares minimization) and time series forecasting methods for predicting demand volume as a function of time period, product price and promotions offered. The testing job included preparation of input data for a variety of use cases, setting up the model independently in excel and checking model vs. test results (demand values, coefficients and regression statistics such as P-value) at various stages within the regression and time series iterations.

 Wrote a generic MIP model for SCA Technologies LLC, unifying the features of various production models. The idea was to minimize model maintenance effort for the future and making various features (constraints and objective function) available under “one roof”.

RESEARCH PROJECTS

• Programmed a linear programming (LP) model for portfolio optimization from a research paper. The quadratic Markowitz function was approximated as a linear function. The benefits of this model were two-fold a) it enabled the real probability distribution of returns (from historical data) to be represented with their implied correlation rather than the usual approximating assumption of the Normal distribution. b) the resulting formulation was convex, obviating the need of any binary variables for this approximation (for non-convex function).

• Wrote a small stochastic programming model for short-term trading on 1-2 assets with a limited number of time periods (up to 5). The objective was to maximize expected returns at the end of the last time period subject to limited downside risk. The downside risk was expressed as the probability of returns being below a certain value (user given). The obvious limitation here is the number of assets and time periods, however with the appropriate approximation this can be extended to scale up. On the other hand it is a very good tool for assets with asymmetric returns such as options. Besides this, it also offers visibility in the form of scenarios which is missing in the traditional Markowitz type formulation.

• Developed a system of algorithms for vehicle routing by using a combination of graph theory (node-packing), genetic algorithm, hill climbing heuristic and simulated annealing. This system is like a “warehouse of combinatorial optimization algorithms”, both stochastic and deterministic. It was built to handle about 900 order points (locations) to be routed with a CPU-time target of 2-3 minutes on an ordinary Pentium machine.

All projects were programmed in C++ and modelled in GAMS (for LP and MIP models).

Conference presentations/ proceedings

• Mokashi S. D. Feb, 2003, Application of OR Modeling in Production planning – A case study from the fast food industry, Workshop on OR Modeling, NITIE (invited talk)

• Mokashi S. D. and Kokossis A. C., 2000, Maximum dispersion algorithms for production-distribution supply chain problems, AIChE Annual meeting

• Mokashi S. D., Kokossis A. C., Sanaka T. and Fujita K, 1999, Collaborative research in logistics optimization, AIChE Annual meeting

• Mokashi S. D. and Kokossis A. C., 1999, The rigorous optimisation of integrated manufacturing operations with a hybrid approach of order graphs and mathematical programming, AIChE Annual meeting

• Mokashi S. D. and Kokossis A. C., 1998, Maximum dispersion algorithm for multi-site delivery scheduling, FOCAPO98

• Mokashi S. D. and Kokossis A. C., 1997, The maximum order tree method: A new approach for the optimal scheduling of product distribution lines, Computers Chem. Engg., 21, S679-S684

Annual Presentations at Process Integration Research Consortium, UMIST

• Optimisation of multiple site schedules, PIRC Annual Meeting 1998.

• Contextual optimisation for scheduling of batch chemical plants, PIRC Annual Meeting 1996.

PUBLICATIONS

• Mokashi S. D. and Kokossis A. C., The maximum order tree algorithm for the optimization of product distribution lines, AIChE J., 2002

• Mokashi S. D. and Kokossis A. C., Application of dispersion algorithms to supply chain optimization, Computers and Chemical Engg. 2003

Contact this candidate