Summary
I am applying to EPFL for a Master's program in electrical and computer engineering specifically in the domain of signal processing and machine learning. I am required to write a statement of purpose (word limit 1000). I am open to any recommendations and changes since I have never written these before.
I've always been incredibly curious to learn about how machines operate and how intelligent they might become. Whether it's a regulator regulating a fan's speed or self-driving cars that can drive almost as effectively as humans, my interest has always been in exploring electronics and information flow withing complex systems. I've always had an analytical mind and a strong interest for mathematics, which has allowed me to solve complex problems using first principles. After carefully considering my passion and inherent skills, I have decided to apply at the Master's program in electrical and computer engineering specifically in the domain of signal processing and machine learning.
My time at XXX university, a reputable university in XXX country, was well spent because it exposed me to a diverse peer group as well as advanced theoretical ideas in engineering and physics. Due to the flexible curriculum espoused by the community I had the freedom to choose subjects which truly interested me. While physics gave me a strong theoretical foundation, engineering gave me the means to apply those ideas in a more practical setting.
I believe one of the most important outcomes in my university was to explore: an opportunity to unlearn and relearn. This started with my genuine enthusiasm with theoretical physics specifically in quantum mechanics. This led me to take up a research-oriented study project under Dr. YYY, where I wanted to further my learning towards the physical implementations of a quantum computer. This truly deepened my understanding towards how quantum bits can be physically harnessed and motivated me to learn more about this field. I also wanted a somewhat different perspective from engineering. To this end, I took up another exploratory venture under Dr. XXX in the field of spintronics. This field fundamentally deviated from traditional electronics: it exploited the discrete nature of the magnetic moment of an electron to store states. My goal was to create a transistor using a magnetic tunnel junction, the spintronic equivalent of a junction diode, through simulations. In addition, I developed a strong interest in circuit design, which led me to enroll in more advanced courses like Analog and Digital VLSI Design.
Another one of the more fascinating courses was digital signal processing, which covered everything from the theoretical underpinnings of filter design to its practical applications. The use of straightforward mathematical principles to successfully extract information from unstructured data greatly intrigued me. In my efforts to further expand my knowledge in this field and apply my circuit design skills, I decided to work under my guide Dr. XXX again in the field of flexible electronics. This was a semester long design-oriented project using organic thin film transistors on RFID tags. I was specifically tasked to design the circuit for the signal processing module. The primary aim of the project was to design a tag that can possibly identify glucose levels from bodily secretions. One of the biggest challenges I faced was to successfully implement frequency shift keying on-chip as well as optimizing the overall cost. My primary learnings were to design signal processing circuits at the transistor level using organic thin film transistors and incorporating them on the RFID tag via simulations.
During my time at XXX university, I had the opportunity to work in cutting edge technology and learn through internships and other part time opportunities. I was selected as a teaching assistant for courses in quantum mechanics as well as electromagnetic theory where I had the opportunity to aid my juniors and at the same time, relearn my concepts. My time at AAA Company, a Swiss-based startup, was very enlightening where I was first introduced to how the field of signal processing and machine learning can be coalesced to detect sleep quality in a contactless fashion. As an intern I was tasked with quantifying signal quality which would be further used in many downstream tasks to optimize their datasets. During this time, I was able to learn basics of exploratory data analysis and engineer features to fit a machine learning model. This experience also contributed to building my passion towards machine learning and AI. My next internship was at Samsung Research where I contributed exclusively on the utility of artificial intelligence in 5G communications. I was able to formulate a strong theoretical background in deep learning since most of my time was spent in going through intensive lectures and completing online courses. My first project was in radio map estimation using generative adversarial networks where I used semantic segmentation to optimize base station positioning. This went on to be published and eventually won the best paper award. My second project was much more ambitious: I was tasked with using reinforcement learning in the field of network slicing to optimize resource allocation. This has also been published in the IEEE ANTS conference. After completing my internships, I went on to become a full-time employee at AAA as a signal processing engineer.
At AAA, I have contributed to a variety of projects using both signal processing and deep learning techniques. My work began in designing an algorithm for distance and human presence detection for FMCW radars and progressed towards more challenging topics such as extracting vital signs data such as heart rate as well as sleep apnoea detection on clinical datasets. Such a range of projects have exposed me to a variety of real-world challenges when it comes to designing the dataset, model architecture and building an efficient inference pipeline which can be deployed on real-time systems. Moreover, I thoroughly enjoyed working with and learning from ETH graduates at AAA. The CTO, ZZZ, helped me develop a hypothesis-driven method of problem solving, which allowed me to motivate a solution rather than merely seek it out.
As a signal processing engineer and data scientist, I believe in making data-driven decisions and EPFL, with its incredible alumni, promises a pulsating culture of innovation. The stimulating environment will allow me to be a part of the cutting edge and eventually becoming a research scientist at some of the top research labs. I believe I could thrive in such an environment and provide value to the institute as well by investing all my time learning and being a part of a great community. The last six years of my life have taught me a lot and instilled in me a competent skill set but at the same time left me with a lot of questions: can artificial neural networks be used to model every aspect of signal processing? Can deep networks predict time series data produced by extremely complex models such as the human brain itself? The master's program at EPFL with the multitude of research opportunities and industry collaborations is the perfect one for me where I am sure I can get the answers.
I am applying to EPFL for a Master's program in electrical and computer engineering specifically in the domain of signal processing and machine learning. I am required to write a statement of purpose (word limit 1000). I am open to any recommendations and changes since I have never written these before.
Statement of Purpose
I've always been incredibly curious to learn about how machines operate and how intelligent they might become. Whether it's a regulator regulating a fan's speed or self-driving cars that can drive almost as effectively as humans, my interest has always been in exploring electronics and information flow withing complex systems. I've always had an analytical mind and a strong interest for mathematics, which has allowed me to solve complex problems using first principles. After carefully considering my passion and inherent skills, I have decided to apply at the Master's program in electrical and computer engineering specifically in the domain of signal processing and machine learning.
My time at XXX university, a reputable university in XXX country, was well spent because it exposed me to a diverse peer group as well as advanced theoretical ideas in engineering and physics. Due to the flexible curriculum espoused by the community I had the freedom to choose subjects which truly interested me. While physics gave me a strong theoretical foundation, engineering gave me the means to apply those ideas in a more practical setting.
I believe one of the most important outcomes in my university was to explore: an opportunity to unlearn and relearn. This started with my genuine enthusiasm with theoretical physics specifically in quantum mechanics. This led me to take up a research-oriented study project under Dr. YYY, where I wanted to further my learning towards the physical implementations of a quantum computer. This truly deepened my understanding towards how quantum bits can be physically harnessed and motivated me to learn more about this field. I also wanted a somewhat different perspective from engineering. To this end, I took up another exploratory venture under Dr. XXX in the field of spintronics. This field fundamentally deviated from traditional electronics: it exploited the discrete nature of the magnetic moment of an electron to store states. My goal was to create a transistor using a magnetic tunnel junction, the spintronic equivalent of a junction diode, through simulations. In addition, I developed a strong interest in circuit design, which led me to enroll in more advanced courses like Analog and Digital VLSI Design.
Another one of the more fascinating courses was digital signal processing, which covered everything from the theoretical underpinnings of filter design to its practical applications. The use of straightforward mathematical principles to successfully extract information from unstructured data greatly intrigued me. In my efforts to further expand my knowledge in this field and apply my circuit design skills, I decided to work under my guide Dr. XXX again in the field of flexible electronics. This was a semester long design-oriented project using organic thin film transistors on RFID tags. I was specifically tasked to design the circuit for the signal processing module. The primary aim of the project was to design a tag that can possibly identify glucose levels from bodily secretions. One of the biggest challenges I faced was to successfully implement frequency shift keying on-chip as well as optimizing the overall cost. My primary learnings were to design signal processing circuits at the transistor level using organic thin film transistors and incorporating them on the RFID tag via simulations.
During my time at XXX university, I had the opportunity to work in cutting edge technology and learn through internships and other part time opportunities. I was selected as a teaching assistant for courses in quantum mechanics as well as electromagnetic theory where I had the opportunity to aid my juniors and at the same time, relearn my concepts. My time at AAA Company, a Swiss-based startup, was very enlightening where I was first introduced to how the field of signal processing and machine learning can be coalesced to detect sleep quality in a contactless fashion. As an intern I was tasked with quantifying signal quality which would be further used in many downstream tasks to optimize their datasets. During this time, I was able to learn basics of exploratory data analysis and engineer features to fit a machine learning model. This experience also contributed to building my passion towards machine learning and AI. My next internship was at Samsung Research where I contributed exclusively on the utility of artificial intelligence in 5G communications. I was able to formulate a strong theoretical background in deep learning since most of my time was spent in going through intensive lectures and completing online courses. My first project was in radio map estimation using generative adversarial networks where I used semantic segmentation to optimize base station positioning. This went on to be published and eventually won the best paper award. My second project was much more ambitious: I was tasked with using reinforcement learning in the field of network slicing to optimize resource allocation. This has also been published in the IEEE ANTS conference. After completing my internships, I went on to become a full-time employee at AAA as a signal processing engineer.
At AAA, I have contributed to a variety of projects using both signal processing and deep learning techniques. My work began in designing an algorithm for distance and human presence detection for FMCW radars and progressed towards more challenging topics such as extracting vital signs data such as heart rate as well as sleep apnoea detection on clinical datasets. Such a range of projects have exposed me to a variety of real-world challenges when it comes to designing the dataset, model architecture and building an efficient inference pipeline which can be deployed on real-time systems. Moreover, I thoroughly enjoyed working with and learning from ETH graduates at AAA. The CTO, ZZZ, helped me develop a hypothesis-driven method of problem solving, which allowed me to motivate a solution rather than merely seek it out.
As a signal processing engineer and data scientist, I believe in making data-driven decisions and EPFL, with its incredible alumni, promises a pulsating culture of innovation. The stimulating environment will allow me to be a part of the cutting edge and eventually becoming a research scientist at some of the top research labs. I believe I could thrive in such an environment and provide value to the institute as well by investing all my time learning and being a part of a great community. The last six years of my life have taught me a lot and instilled in me a competent skill set but at the same time left me with a lot of questions: can artificial neural networks be used to model every aspect of signal processing? Can deep networks predict time series data produced by extremely complex models such as the human brain itself? The master's program at EPFL with the multitude of research opportunities and industry collaborations is the perfect one for me where I am sure I can get the answers.
