The monthly LGO Alumni Newsletter goes out by email to all alumni; archived issues are posted on this page. If you would like to change your contact information, please email Josh Jacobs.
Thomas A. Roemer
April passed by in a whirlwind, and a reluctant spring seems finally to be arriving in Boston. Our graduating students are finalizing their thesis and starting to prepare for the next big steps in their lives even as the class of 2018 starts taking its shape. We are finalizing the last few admission decisions, so I can't provide final statistics on the LGO '18s, but everything looks great and we're excited about the incoming class.
Meeting the new students during AdMIT Day was certainly one of the highlights for April. A very successful reception at Aeronaut Brewing included a “factory” tour by the founders (all MIT Ph.D.s) and helpe emphasize how vibrant our city is and what pivotal role MIT has played in shaping its unique culture and diversity.
For May, I'm especially looking forward to the annual LGO Alumni Conference in Charleston and hope to see many of you there. Click on the conference website for more information about the agenda and speakers, how to register and a list of alumni who have already signed up to attend. If you have any questions regarding the conference, please contact Josh Jacobs.
All the best, and I look forward to seeing you in Charleston!
Thomas A. Roemer
Executive Director, MIT Leaders for Global Operations Program
Are you planning to attend the MIT Sloan reunion or expect to be on campus for MIT Commencement on Friday, June 3? Please join an informal LGO alumni meetup at 5 p.m. at Firebrand Saints (One Broadway in Kendall Square) immediately following the LGO ’16 graduation party starting at 3 p.m. there. LGO is buying the first pitcher. Please let us know at email@example.com if you plan to stop by.
Shanahan, a member of LGO's second class, is now SVP, supply chain and operations. Read more..
German firm is a leader in electromechanical control systems for electrical power transformers. Read more..
The Alumni Newsletter includes periodic first-person accounts by students who have benefited from an LGO Alumni Scholarship.
Amy Gobel is a first-year LGO student in engineering systems with a focus on supply chains and manufacturing systems. She graduated from Princeton in 2012 with a bachelor’s degree in geosciences and spent her time before LGO working for ENVIRON, an environmental consulting company focused on investigating and remediating contaminated properties.
I've known since high school that I wanted to tackle environmental problems at a global scale, but until I came to the LGO program, I didn’t know what that career would look like.
I studied geosciences in college because I wanted to understand how humans interacted with and impacted natural systems. After college, I joined an environmental consulting company, ENVIRON, so I could start to mitigate those impacts. My group specialized in human health risk assessment: evaluating the potential health impacts from chemical spills and industrial waste to determine the best way to clean up the site. I enjoyed the work, but I realized that I was getting to the problem too late. If I wanted to make a bigger impact, I needed to look upstream and prevent the spills and waste from happening in the first place.
After that realization, I began talking to faculty mentors about how I could make a transition to industry. Some recommended a business degree so I could understand the financial and managerial levers needed to make change. Some recommended an engineering degree so that I could make informed decisions about technical processes. Then I found out about LGO, which allowed me to do both—it seemed like a perfect fit.
Of course, doubts started to creep in when I began the application process. I had too little experience. I didn’t have an engineering degree. But when I got the call from Thomas and found out that I was both accepted and also awarded an Alumni Fellowship in addition to the already generous LGO Fellowship, I was thrilled to know that LGO thought I was a good fit, too.
With this extraordinary support, I have been able to join a program that has exceeded my high expectations. Through the academic curriculum, the plant treks, and the continual education from my classmates and friends, I've begun to see how I can apply my sustainability perspective to manufacturing challenges to help industry while helping the environment. In fact, my internship this summer at Quest Diagnostics will involve reducing reagent waste in their diagnostic processes, which exemplifies this goal of cutting their costs and shrinking their footprint, and I know that the LGO program has prepared me to make the best use of this opportunity.
— Amy Gobel '17
Viju Menon, LGO '94
Education before MIT: University of Michigan at Ann Arbor (Computer Science and Engineering)
While at LGO: MBA and SM in Electrical Engineering and Computer Science
Current: Senior Vice President, Global Supply Chain, Verizon
Viju came to LGO from Intel and subsequently continued his 19-year Intel career, distinguished by key leadership roles such as factory manager and subsequently head of worldwide supply planning operations. In 2010, Viju joined Verizon Wireless as vice president, supply chain. In 2013, he was named senior vice president, global supply chain, at Verizon Communications.
“LGO proved to be an amazing inflection point in shaping my world view, developing my leadership skills and influencing my subsequent career trajectory," Viju says. "Even after 20 years, I still draw from my LGO toolkit when it comes to thinking big and leading transformational change. I am a big LGO believer, and it was a no-brainer for me to sign up Verizon as an LGO partner.”
This month will feature two projects involved with different aspect of machine learning with electronic technology firms. Naomi Arnold worked with machine learning algorithms to drive yield improvement opportunities at SanDisk. Mario Orozco explored machine learning as one alternative for Dell to drive customer satisfaction and business performance.
If you have companies or areas of research you are interested in having highlighted in the monthly news, please contact Ted Equi.
Problem: In the semiconductor industry where the technology continues to grow in complexity and manufacturing costs, it is becoming increasingly important to drive cost savings by screening out defective die upstream. SanDisk would like to predict defective wafer die before in-line testing occurs at their Shanghai assembly facility (SDSS). SDSS runs a costly test process when wafers arrive called Known Good Die (KGD). After assembly, the final memory test (MT) exposes additional defects. The project hypothesis is that some wafers can skip KGD and avoid MT die failures given indicators in the upstream parametric data.
Project Goals: The primary goal of the project is to build a statistical prediction model to facilitate operational improvements across two global manufacturing locations. A prediction model could improve the accuracy of wafer and die sorting, resulting in decreased assembly costs. It could also enable coordination with the wafer fab facility to implement root-cause fixes to further drive cost savings.
Approach: The scope of the project is one high-volume product line, an off-line statistical model using recent historical mass production data, and experimentation with machine learning algorithms. The project involved partnerships with experts across the company from test, quality and process engineers in China, U.S. and Japan as well as the expertise of data scientists and data owners. The prediction model progressed in three phases. The first two phases established methodologies for data refining of hundreds of predictor variables, selecting top defect soft bins (binary response), sampling and evaluation metrics (confusion matrices). The methodologies learned from the KGD model were applied to the MT model.
Results: The first phase was a successful proof-of-concept wafer-level Random Forest model that predicted a new KGD defect. The second phase was a successful die-level prediction model that predicted top KGD defects. Tree-based algorithms such as Random Forest performed best. For the last phase, a large dataset of 1.5 months with over 1,000 predictor variables was included. CART and Random Forest were the main algorithms utilized, but due to computational performance and time limitations, the MT prediction model remains unproven.
Conclusions: There exists a wafer sort improvement opportunity to drive cost savings by predicting KGD results, yet more work is needed to conclude the same for MT prediction. Key findings include: the importance of model computational performance for big data problems; the necessity of a living model that stays accurate over time to meet operational needs; and an evaluation methodology based on business requirements is proposed (ratio of overkill to correct predictions) to determine model probability acceptance criteria. Lastly, this project provided a case study for a high-level strategy of assessing big data and advanced analytics applications to improve semiconductor manufacturing.
Mario Orozco Gabriel
Dell is committed to delivering reliable, secure and manageable devices which have to be built with ever-evolving technology to adapt to changing needs. “Intelligence” is a common theme among projects to meet new requirements across all areas at Dell: Operations and Client Services, Enterprise Solutions, Services, and Software. However, can “intelligence” be applied at Dell, and if so, where and how?
Goals: There were three primary goals for this internship: (1) explore AI technologies with the purpose of finding applicable ones that are a fit with Dell (business optimization and internal applications), provide further recommendations of potential areas of application of AI technologies to Dell, and (3) document current and future state of a selected process and develop models to quantify such improvements.
Approach: The approach to achieve the main goals was to first research the field of AI. With the results, we developed a technology taxonomy. Once this was defined, we brainstormed 30+ potential applications at Dell, which resulted in the areas of security, serviceability, manageability and productivity. Next, to prioritize the concepts, we applied a framework that took into account tech readiness, financial opportunities, IP potential, Dell fit and reputation. The considered technologies for different applications included machine learning, deep learning, natural language processing, and speech and image recognition.
A detailed business and technology case was done for the top four concepts, and one was selected to further analyze data: self-management and healing (computer hardware failure prevention). Currently, computer failures are treated post-event, which creates a hassle for the user and has an important monetarily impact on the service provider. However, the recent launch of the Dell Data Vault—a software application o capture hundreds of variables from the hardware components, types of alerts, and failure types in a computer or laptop—enables capture relevant data. With this data and utilizing a data science approach, we analyzed it and have the potential to prevent hardware failures by warning the user or even self-correcting.
Results: Four concepts were identified with direct impact in the millions of dollars, but two were recommended to pursue directly due to higher impact and technology readiness. Currently, analysis through machine learning is being done to corroborate the efficacy in preventing computer hardware failure.
Conclusions: Machine learning technologies can be applicable across organizations and present great opportunities in cost reduction/avoidance, delighting customers and creating new claims for products. Constant monitoring is essential due to the rapidly changing technologies and emerging players.
The Alumni page of the LGO website explains how your donation to one of our alumni gift funds will benefit current and future students.