 Good afternoon everyone and welcome to the finals of this year's Carnegie Mellon University three-minute thesis competition. My name's Keith Webster I'm the Helen and Henry Posner Junior Dean of University Libraries. I'll be your host this afternoon. After a two-year hiatus due to the global pandemic we're thrilled to be back for the seventh year of hosting three-minute thesis at CMU. We continue to attract an impressive array of doctoral students willing to share their research in an informative and entertaining manner. It's wonderful to see so many of you with us in the Kresge Auditorium today. In addition thanks to the support of the CMU Alumni Association we are live streaming the competition. The three-minute thesis competition was developed by my former university the University of Queensland in Australia. We started the event in 2008 as an in-house opportunity. It grew now to be an international competition being held in hundreds of universities around the world. The premise of the three-minute thesis is simple. Speakers have, guess what, three minutes to present a compelling aeration on their thesis and its significance. It's not an exercise in trivializing or dumbing down research but challenges students to consolidate their ideas and research discoveries so they can be presented concisely to a non-specialist audience. In bringing the 3MT challenge to CMU I had no idea of the likely response. Over seven years we've had almost 500 participants from across the university and judges representing faculty and staff from colleges, schools and the university libraries. University leaders and representatives of the research student body have also participated enthusiastically. I acknowledge everyone's contribution to 3MT with thanks. Thirty-eight students from every college participated in this year's competition and we have three colleges represented in this year's finals. I'm pleased to introduce to you this evening's judges. They're all seated in the front row wondering why an earth they're here. Mary Ellen Poole, the Stanley and Marcia Gumberg Dean of the College of Fine Arts and steward of this building. Mary Ellen joined during the pandemic and this is certainly her first outing at this event. Scott Mori, Vice President of University Advancement. Teresa Trombetta, Assistant Vice President for Alumni and Constituent Engagement. Chris Stengel, President of the Carnegie Mellon Alumni Association and Krista Bond, Vice President of Academic Affairs in the Graduate Student Assembly. herself a PhD candidate in Cognitive Neuroscience. Thank you all for your participation. The rules are draconian according to Mary Ellen. She just pointed out to look rather severe. I will just say we developed these in Australia. Go with the flow. The rules are actually quite straightforward. Speakers are limited to three minutes and the competition rules required the disqualification of anyone who continues beyond that time. If they choose, speakers can view a countdown timer during their presentation. Ryan Splendor, one of our faculty in the libraries, will maintain the timer and will wave a 30 second warning and sound a bell at the end of their allotted time. Speakers are permitted to display a single static slide but may not use any other media props, rap, singing, dancing or anything of that sort. But they've been through it before they know what they're doing. Our judges will consider three broad criteria in arriving at their decision. These are comprehension, engagement and communication. In just a moment I will introduce each speaker in turn and invite them to come to the stage to deliver their presentation. We will go through this pretty briskly once I stop giving my six minute thesis. Once the speakers have completed their presentations, the judges will retire to determine today's winners. Our judges will select three winners who will receive awards as shown on the screen. In addition to the handsome prizes, they have already won in qualifying for these championship finals. In addition, the audience here in Kresge Auditorium will have an opportunity to vote for the People's Choice Award. At the end of the presentations we will collect your ballot paper which is also the program for the event. Please vote for the candidate of your choice. Those worried about cyber security will be relieved to know that we've adopted a hack-proof paper-based medium for that purpose. And we are delighted once again to partner with the CMU Alumni Association to award the Alumni Choice Award. This $500 travel stipend will be awarded to the participant who receives the most votes from CMU Alumni watching the live stream on the Alumni Association's website. We'll remind you how to vote at the end of the presentations and the winner of the Alumni Choice Award will be announced at the end of this evening's event along with other winners. Before introducing our first presenter I would like to thank the many people who have made this event possible. Our finalists have arrived here from eight heats which together attracted students from every college in the university as I mentioned already. I'm grateful to everyone who participated as a contestant. I'm grateful to those who served as judges at the heats and I acknowledge the phenomenal work of my colleagues in the university libraries especially Erica, Ryan, Sarah, Sonja, Morgan, Shannon R, Shannon B, Andy and Heidi. One final request of the audience here please check that your cell phones are muted on silent on airplane mode whatever works for you. And with that I am delighted to welcome our first presenter Emma Benjaminson who will present her three minute thesis stealing magnetic robots with MRI technology. Magnetic swimming robots have the potential to revolutionize targeted drug delivery and medicine. By delivering small drug loads to a target site in a patient we can reduce unwanted side effects and maximize the therapeutic benefits of the drugs that we administer. Researchers have shown that it is possible to use a hospital's existing magnetic resonance imaging or MRI technology to perform targeted drug delivery with magnetic robots. The benefit to using an MRI machine is that you can alternately image the patient to determine where the robot is located and control the robot using a combination of magnetic field gradients to pull the robot along three axes. However researchers have shown that exposing patients to magnetic fields that alternate rapidly between imaging and control can be harmful. Data shows that patients who are exposed to magnetic fields that rapidly alternate at frequencies of over 100 hertz can begin to experience peripheral nerve stimulation the symptoms of which can range from a mild tingling in the fingertips to a heart arrhythmia. And unfortunately we might have to operate at these frequencies if we're going to navigate complex regions of the body at high speed with small robots. So to overcome this challenge I developed a novel approach to perform imaging and control simultaneously. Normally an MRI machine will use magnetic field gradients along three axes to capture 3D scans of the human body but it's also possible to capture 2D slices leaving the third magnetic field gradient available for use. Now a magnetic field gradient along just one axis might not seem like enough of a control input to guide a robot through a 2D plane but if you observe the field lines generated by such a gradient you can see that they have some curvature in two dimensions. The benefit of this curvature is that a magnetic robot could essentially ski along one of these lines and by alternating the direction of the field now the robot can slalom through the plane. Finally by incorporating a switching time optimization algorithm we can intelligently determine when to change the direction of the field to steer a robot to a target site in a 2D plane with one control input. The benefit to this approach is that the rate at which we need to change the direction of the field is much less than 100 Hertz thereby eliminating the risk of exposing the patient to peripheral nerve stimulation. Another benefit is that by using only one gradient for steering we are left with two additional gradients to perform simultaneous imaging. This technique allows us to adapt the hospital's existing MRI technology to better perform targeted drug delivery in humans. Thank you Emma. Our second presenter is Edithi Choudhury whose title is prediction and application of linguistic insights. Okay. Who here has had to learn a new language say for immigration purposes? Millions of people are learning new languages every single day for job, tourism, immigration. In fact, Diolingo, a popular language learning app has shown that it can even help refugees to rebuild their lives. So clearly there is a need for good learning materials and tools which people can have easy access to. And language learning is not only important for people to achieve their own goals but it's also important for language as a whole to survive. For example, a recent report from UNESCO shows that there are 191 languages on the verge of extinction in India alone and the reason for that the primacy given to English as the medium of instruction in schools which causes many local languages to be ignored. So why not simply add these languages to Diolingo? It's not as easy as one may think because you need subject experts to come in, sit for days, maybe even months to create the curriculum, find relevant examples, it's time consuming. And by the way there are 7,000 such languages in the world today and for some of these languages the experts are even inaccessible. So in my PhD I have explored techniques from machine learning, natural language processing and even linguistics to help automate some of the processes of a curriculum extraction. Existing methods can already take some English sentences and identify basic linguistic information like what is a proper noun, what is a pronoun in that sentence. In my PhD I have improved upon these methods to identify such basic information for potentially any language. So on the top I have an example of Marathi which is my native tongue. Now this basic information of proper noun, pronoun is useful but is not sufficient because language is complex. You also need to know how to arrange the words appropriately, what if the speaker is of a different gender, how to modify the words. So for each such linguistic question I formalize it into a machine learning task, I learn an algorithm for it and then extract rules which can help illustrate each phenomena. So each component of the curriculum is now automated to some extent. Now rules alone could get overwhelming so I also extract automatically some illustrative examples to highlight each phenomena. And so each process is now completely automated and in fact this curriculum is now being tested in actual North American schools which teach Indian languages. So if you also give me text in your language, I can create a first pass curriculum for you. It won't be perfect but it'll be a good start. And again the whole goal is not to replace a teacher but rather to assist them in their teaching process so that they can focus on the creative aspect of teaching. That's all from my side, thank you. Thank you for that presentation, I was conscious I didn't give the judge as much time between the first and second presenter to take notes so I'm just pausing for a moment. Our third presenter is Gaurav Balakrishnan whose title is gelatin-based diagnostic edible electronics. $136 billion that is a staggering annual cost worn by patients and the health care system to treat, diagnose and monitor gastrointestinal diseases such as Crohn's disease, celiac disease and several types of cancer. Patients suffering from such conditions often require frequent visual monitoring of their digestive tracts and retrieval of tissue samples via an endoscope, a long tubular probe. For context there are over 50 million endoscopies performed in the United States each year and a patient with chronic gastrointestinal illnesses might need an endoscopy once every two to three months. As a result there's considerable, as you can imagine, endoscopies are painful procedures that can cause discomfort to a patient. As a result there's considerable interest in designing electronic capsules to replace endoscopies. There are over 50, sorry I apologize, this idea is validated by a $550 million market for capsule endoscopies in the United States today. However capsule endoscopies currently on the market are composed of the same type of electronics that make up your phones and your computers. That is they're rigid and non-degradable. This creates the risk of obstruction or tissue damage if they were to get stuck as it's being swallowed. My research leverages the intersection of material science, biomedical engineering and electronics to design soft and degradable electronic capsules. To do this we start by designing a class of Jell-O or gelatin-based materials with highly tunable mechanical and degradation characteristics. Next we develop techniques to transfer electronics typically fabricated on rigid silicon wafers onto our soft gelatin materials. And lastly we picked a specific medical application as a first test of our edible electronics platform. We design sensors that can quantitatively measure how closely packed the cells of the digestive tract lining are. In a healthy patient we expect these cells to be very tightly packed whereas multiple gastrointestinal diseases can induce an increased intercellular spacing. We have tested our devices with both artificial and animal-based tissue models and we are currently working with doctors at the University of Pittsburgh to test these in humans in the short term. I believe that the exciting research I'm showing you today will serve as a blueprint for the next generation of pain-free low profile and most importantly patient-centric edible electronics for diagnosis and therapy. Thank you. Our fourth presenter is Piumi Vijisakara whose title is engineering rotating mini lung tissue for combating respiratory infection. Human lung is a finely tuned gas exchanging organ. Even though this may sound surprising lungs are body's largest interface with the outside environment being approximately the size of a full record board court. When we inhale the external environmental factors such as wires, bacteria, smoke, pollutants they first interact with these tube-like airways in our lung. The coronavirus disease that has affected millions of people worldwide is a great example of this where the virus first binds to the receptors in our airways before initiating the infection. If you take a closer look the lumen of this tube-like airway that we call the apical surface is composed of a group of cells including ciliated cells with thin hair like cilia constantly beating and gulped cells with little sacks producing mucus. So whenever a foreign particle enters into our airway they get trapped in this produced mucus and with cilia constantly beating push these particles upward in our airway towards the throat so we can speed them out. Considering all these facts the human airway or the lung is a constant battlefield between our body and the outside environment so having a good model to study these interactions is extremely important. There are current airway or lung models however they're highly undesirable. The reason is they have a disoriented apical surface where it's hidden inside the tissue unlike in our human lung they're highly inaccessible to the external environment factors that we want to test when modeling infection. So to address all these challenges in my thesis I have engineered this mini lung tissue in a dish with the apical surface facing outside that can directly interact with the respiratory pathogens and pollutants. As you can see in this engineered mini lung tissue the ciliated cells are selectively present on the outside constantly beating which also gives them this unique ability to rotate when embedded in a supporting material. Since this rotation of motion correlates to individual cilia beating this can be used to transform current cumbersome nanoscale special equipment demanding cilia beating assessment methods into more convenient micro-scale methods. Being similar to exactly how our lung would function this engineered mini lung tissue can also be used to study lung biology lung diseases targeted therapies in a very accurate and efficient manner. In particular as you're still living in a pandemic having an accurate lung model is critical to respond to the virus immediately and effectively. So if we ever hit by another pandemic hopefully not we are better prepared thank you. Our fifth presenter is Amaranth Kara whose title is 3d printing of high-temperature melting metals why and how. Let's just do a quick survey in this room. Growing up how many of us have watched Star Wars or any other sci-fi films and wanted to become an astronaut one day? That's a great number. How many of us if ever so vaguely remember the 2003 Colombian space shuttle disaster that upon re-entry into the earth's atmosphere shattered into a million pieces taking the life of those astronauts who made their childhood dream a reality? Now although there are multiple reasons for this crash one of the primary reasons is the inability of the material to maintain strength at such high temperatures. Researchers ever since have been trying to solve this problem although with a little success either the choice of materials are not great enough or it takes a lot of taxpayers money. We at our lab believe that we have the right mix of new age tech and proper materials to make sure that the history does not repeat itself. Those materials that not only withstand high temperature strength but also are manufactured at a lower cost with the help of metal 3d printing. Now metal 3d printing unlike traditional routes of manufacturing uses a complex design given by a computer and a suitable technique. In our case that suitable technique is called the directed energy deposition. Now as the name suggests it directs a laser beam on the metal surface melting the metal surface and the powder thereby getting the layer by layer structure. You can think of it as building something from Legos. Now throughout history tungsten has been there for a while. People did use tungsten although because of its high temperature melting and also its crackability it's been an issue to use in space applications. However with the advent of 3d technologies tungsten and tungsten like materials like tantalum have shown some initial good steps. With the help of energy conservation relationships we believe that it is not tungsten and tantalum but a mixture of tungsten tantalum and rhenium that could solve this problem. Our 3d simulations have shown us that this mixture gives us not only a very good strength high temperature material but also at a lower cost that could be manufactured. To test this out and to test our technique out we try to print tungsten tantalum and we believe we are a step in the right direction. We hope that in the future we become the inspiration for the next gen scientists and we hope that the loss of lives of these scientists is not gone in vain. We think that the the future of space travel is safe and economic with our research. Thank you. Our sixth presenter is Susie Lee whose title is adopting smart neighborhood surfaces as critical to climate change, human health and social equity. Did you know we're losing our green land to more and more pavement every year? On average almost 70% of our cities are paved now. As a result increasingly cities are facing extreme urban heat and flooding problems. With around 1,300 heat related deaths in the U.S. every year and over 4.3 million losses for homes due to flooding damage. Now it might say getting more paved surface like streets and sidewalks is inevitable for urban growth but are there smarter choices for urban development? Luckily there are if people start to pay more attention to the surface reflectivity and permeability. Reflectivity is largely related to the color of the surface. Low reflectivity causes more solar heat to be absorbed leads to an increase in surface and air temperature which results in an increase in surrounded building air conditioning load powered by the polluting power plants which is a major contribution to climate change. And this increase in temperature is consistent with the lack of tree coverage in those red line neighborhoods as shown on the map indicating social inequity. On the other hand low permeability leads to an increase in storm water runoff causes water contamination and human health issues. So by choosing a smarter surface choice those errors on the top can be flipped into reductions but current data on city surface performance is not easily accessible or comparable. Most research have been focusing on either one type of the surface or one aspect of this issue heat or water but they should be considered both simultaneously. So to fill this gap I've developed a smart surface taxonomy with performance indicators in surface temperature, rainwater retention and carbon reduction. There are 50 type of surface in the smart surface taxonomy covers three categories roofs streets and sidewalks and parking lots ranging from traditional black impervious to light colored or green surface with various levels of permeability. So let's take look at the roof category. The green roof on this bar chart is 80 degrees cooler than the conventional black roof on a hot summer day and like I said before this difference in temperature is showing a strong correlation with the lack of tree coverage in the communities of color. So there are smarter ways for the cities to grow and become more equitable. The question is why are the policymakers still acting so slow on adopting these strategies? That's because most decisions are made economically driven instead of sustainability. So in this research a decision making dashboard will be developed to help them to evaluate some other potential benefits like health benefits or flooding cost reduction so that more informed decisions on city surface choices can be made to help mitigate climate change and improve human health and social equity. Thank you. Our seventh presenter is Mohammed Ayaz Masoud whose title is laterally actuating phase change nano relay for non-volatile memory operation. Your cloud storage is full to upload this file by more space. Don't you hate that notification but to be honest the big cloud companies require space and energy to store your data and believe it or not more energy is lost due to leakage than what is required to upload store and download your files. With my research we should be able to reduce that leakage energy to an absolute zero. Let me explain how. At the core of any electronic device we have transistors. A transistor is a channel for electron to flow from a source to a drain controlled by a gate. Now 20 years ago the distance between the source and the drain was hundreds of nanometers. Today there are only five to seven nanometers apart barely 50 atoms so when you are trying to turn it off some electrons always find their way across hopping on those few atoms that is leakage and as you have billions of transistors in an area of just one centimeter square the total leakage adds up significantly contributing to a warmer planet. So what could be solution? Well if we could remove those few atoms creating a gap electrons cannot jump across anymore hence we get zero leakage but in that case how do you turn it on? You'd have to mechanically move the source towards the drain to make a physical contact on off on off just like a light switch an electromechanical switch and to be used as memory it has to be stable in both on and off states but to make a five nanometer size stable controllable mechanically moving structure is almost impossible. So in our group we redefine this architecture we use germanium telluride GET a phase change material we can change the volume of this material using electrical pulses and that change is rapid reversible and stable. We put a thin metal layer on top of GET so that when it expands it connects a source and drain creating a channel for electron to flow and as it shrinks back the gaps reappear perfect zero leakage. For my thesis I have used this architecture to make nanoscale devices these devices are comparable in size to today's transistors I have made the devices vertical to further reduce their footprint I developed nano fabrication techniques to deposit and pattern multiple layers of vertical thin films imagine a vertical stack of papers now imagine that in a nanoscale in the off state these device these device indeed has zero leakage and you can turn it on with lower voltage than conventional transistors if you look at that pie chart we're not going for a slice here we're going for the entire pie these devices have application in space technology cloud memory quantum computing and artificial intelligence the time is a buzzword zero leakage nanoscale memory has an application in all realm of technology so that the next time you upload a photo of your loved ones only your heart gets warmer not this planet thanks for listening. Our eighth presenter is Satvik Divy whose title is role of latches and latch mediated spring actuation systems for high acceleration movements and small-skiller robots. A grasshopper can jump to about 20 times its body size if I as a human would have to match that in a single leap I would have to cover more than a third of a football field yeah that's definitely not happening and in fact that's not the only fascinating thing about these grass hoppers and other such small insects we often use the phrase in the blink of an eye to describe something that's really fast right well the blink of a human eye on an average can take about 200 300 milliseconds long the jump sequence that powers the jumping movements of these small insects that happens in a millisecond that's orders of magnitude faster in fact that's not just fast at this point that is ultra fast right so imagine if we can take inspiration from these small insects and build a small one gram jumping robot we'd ideally want it to be able to do three things right one of course jump in a given direction number two have the ability in spite of its size to be able to control that jump distance right maybe instead of jumping a meter jumps have a meter or maybe just jump two meters right in number three we'd want it to be able to reliably do this in different environments because ultimately we want these devices deployed out to the real world right so imagine if this robot was out there on the cut right there's different environments there there is grass there is sand there's also the pavement which is as hard as a rock there was snow a couple of days ago right so there's a whole different environment so the robot would have to adapt its jump to be able to navigate right which is similar to humans as well like me jumping on the stage is going to be different from let's say jumping off of a diving board by the pool or jumping on a trampoline so how do we do all of these things while keeping in mind the ultra fast nature of these movements that's where my research steps in it does three important things number one it helps establish a framework that's useful for analyzing these ultra fast movements in biology and help translate these principles into the realm of engineering especially small robots and number two by building on this framework we can show how to control these ultra fast movements and number three by further using this framework we can design these small resource constrained robots in such a way that they can adapt their jump based on the environment they're jumping from and we fundamentally are able to do this by relying on the design architecture of these insect inspired mechanisms and by addressing these three key research areas what we're able to do is push the boundaries of our current capabilities when it comes to making these small insect sized resource constrained robots where we are able to overcome not only the size limitations but also the time limitations of operation last but not the least the contribution of my research may be a small step in humanity's knowledge but it is a giant leap when it comes to insect robots thank you so thank you to all of our presenters congratulations on making it to the finals thank you for wonderful presentations please join me in thanking our eight finalists as I mentioned at the beginning there are two audience voting opportunities for those of you watching via live stream a reminder that if you intend to vote for the alumni choice award you may do so now via the alumni association website for those of you here in kresge theater please vote on the back of your program and my colleagues will be around to collect your ballots shortly our five judges now will now depart to deliberate and we'll be back as soon as we can with the judges decisions in the meantime some entertainment I hope this podcast dives into the university's archive of recorded oral histories to showcase the people okay thank you for your patience this was tough this year again thanks to our eight finalists for their great presentations thanks to our judges for their support so conveniently we have five awards five judges so each judge will award something relevant to their role so awarding tonight's people's choice award I'm pleased to introduce Krista Bond and the people's choice award is presented to Mohammed Ayaz Massoud alumni choice award voted on by our live stream audience will be presented by Chris Dengle and the alumni choice award is presented to Amranth Kara I didn't realize they were rearranging themselves tonight's third place award will be presented by Teresa Trambetta and it is awarded to Emma Benchmanson tonight's second place award is presented by Mary Ellen Poole and is awarded to Susie Lee finally the 2022 three-minute thesis champion is awarded by Scott Moray and the first place award is presented to Piyumi Wigisakara so thank you again to all of our presenters and judges to you the in-person audience to our live stream audience have a good evening enjoy carnival next week look out for the many libraries events that will be taking place then have a great evening