Hi, I am applying to EPFL MSc in CS Program. Any criticism, suggestions, or feedback would be much appreciated.
This is how EPFL describes SOP: The statement of purpose should not exceed 1000 words. You are required to describe your academic background and your career strategy. Please be precise about the objectives you wish to reach through your studies at EPFL.
SOP
As an AI Engineer at ABC, I am responsible for developing AI solutions for downstream tasks and the integration of models and AI applications with governance frameworks for explainability, bias, and safety monitoring. In 2023, during the development of a financial report generation model, we discovered that fine-tuning alone led to suboptimal performance on reasoning tasks. When considering how to address this challenge, I learned about chain-of-thought prompting and employed this technique, resulting in a 50% performance increase. This experience highlighted the importance of in-context reasoning in large language models (LLMs) and sparked my curiosity about reasoning and in-context learning in LLMs.
I aspire to transition into a research engineer role at ABC Research, working on projects focused on in-context learning and enhanced reasoning in LLMs. Reasoning plays an important role in LLMs, contributing significantly to their reliability and trustworthiness. While LLMs have shown their capacity to answer questions through in-context reasoning, in-context reasoning can be sensitive to additional content, leading to degraded reasoning and performance. As a result, I aim to develop algorithms and methods for robust reasoning.
Through EPFL's MSc in Computer Science program, I hope to expand my theoretical foundation in machine learning and gain research skills to explore ways to enhance reasoning in LLMs, thereby enhancing performance and robustness. Core courses such as Modern Natural Language Processing would provide the foundation to implement NLP models, understand their failure modes, and develop mitigation techniques for these issues. In the same vein, elective courses such as Deep Learning will help me interpret the performance of deep learning models, analyze their limitations, and address them.
I am particularly interested in exploring the possibility of conducting research with Prof. Antoine Bosselut's group. His work in the development of algorithms and frameworks to improve reasoning in LMs deeply interests me. The opportunity to work with Professor Bosselut, other CS faculty members, and my peers will be a valuable resource in my academic and professional development. Thus, I am certain the MSc CS program at EPFL will equip me with the theoretical foundation and research skills to develop algorithms and techniques to enhance reasoning and learning capabilities in LLMs.
At XYZ University of Technology, I majored in Electrical Engineering, gaining a strong foundation in mathematics, numerical methods, control systems, and MATLAB. For my final year project, under the guidance of Prof. AAJJ, I developed an approach for detecting islanding in distributed generation systems using ANOVA and autocorrelation, which was awarded first place in the department-wide submission. Additionally, in my final semester, as a research intern under the guidance of Professor SSGG and Professor DDBBi, I investigated language models and generative models to improve image-text alignment and language comprehension in text-to-image models. This work culminated in a review paper that was published at IEEE 2023. These experiences gave me an understanding of statistical analysis, NLP, and generative modeling and exposure to a rigorous research environment, preparing me to tackle course concepts and collaborate with faculty members on research projects at EPFL.
After graduating, in 2021 I joined the YCX as a Machine learning Engineer where I engaged in research and development of models and algorithms for predictive maintenance, energy generation optimization, and demand forecasting. Notably, I collaborated with two machine learning engineers and developed a deep learning-based time-to-event model for predictive maintenance in a Japanese bio-fuel power plant, which resulted in a 22% reduction in unplanned downtime. Currently, at ABC, I have furthered my skills by developing a system for automated product recommendation and marketing message generation. I utilized in-context learning with a Llama 2 model for personalized product recommendations and fine-tuned it with QLoRA for personalized marketing message generation based on product recommendations and demographic profiles. Through these professional experiences, I enhanced my skills in model development and deployment and learned how to collaborate and confidently communicate my work, preparing me for future research opportunities.
My academic and professional experiences have provided me with the knowledge, skills, and exposure required to pursue a rigorous multidisciplinary program and contribute to enhancing the reasoning capabilities of LLMs. I am confident that an education at EPFL will allow me to grow as a research engineer and contribute to research in this field.
This is how EPFL describes SOP: The statement of purpose should not exceed 1000 words. You are required to describe your academic background and your career strategy. Please be precise about the objectives you wish to reach through your studies at EPFL.
SOP
As an AI Engineer at ABC, I am responsible for developing AI solutions for downstream tasks and the integration of models and AI applications with governance frameworks for explainability, bias, and safety monitoring. In 2023, during the development of a financial report generation model, we discovered that fine-tuning alone led to suboptimal performance on reasoning tasks. When considering how to address this challenge, I learned about chain-of-thought prompting and employed this technique, resulting in a 50% performance increase. This experience highlighted the importance of in-context reasoning in large language models (LLMs) and sparked my curiosity about reasoning and in-context learning in LLMs.
I aspire to transition into a research engineer role at ABC Research, working on projects focused on in-context learning and enhanced reasoning in LLMs. Reasoning plays an important role in LLMs, contributing significantly to their reliability and trustworthiness. While LLMs have shown their capacity to answer questions through in-context reasoning, in-context reasoning can be sensitive to additional content, leading to degraded reasoning and performance. As a result, I aim to develop algorithms and methods for robust reasoning.
Through EPFL's MSc in Computer Science program, I hope to expand my theoretical foundation in machine learning and gain research skills to explore ways to enhance reasoning in LLMs, thereby enhancing performance and robustness. Core courses such as Modern Natural Language Processing would provide the foundation to implement NLP models, understand their failure modes, and develop mitigation techniques for these issues. In the same vein, elective courses such as Deep Learning will help me interpret the performance of deep learning models, analyze their limitations, and address them.
I am particularly interested in exploring the possibility of conducting research with Prof. Antoine Bosselut's group. His work in the development of algorithms and frameworks to improve reasoning in LMs deeply interests me. The opportunity to work with Professor Bosselut, other CS faculty members, and my peers will be a valuable resource in my academic and professional development. Thus, I am certain the MSc CS program at EPFL will equip me with the theoretical foundation and research skills to develop algorithms and techniques to enhance reasoning and learning capabilities in LLMs.
At XYZ University of Technology, I majored in Electrical Engineering, gaining a strong foundation in mathematics, numerical methods, control systems, and MATLAB. For my final year project, under the guidance of Prof. AAJJ, I developed an approach for detecting islanding in distributed generation systems using ANOVA and autocorrelation, which was awarded first place in the department-wide submission. Additionally, in my final semester, as a research intern under the guidance of Professor SSGG and Professor DDBBi, I investigated language models and generative models to improve image-text alignment and language comprehension in text-to-image models. This work culminated in a review paper that was published at IEEE 2023. These experiences gave me an understanding of statistical analysis, NLP, and generative modeling and exposure to a rigorous research environment, preparing me to tackle course concepts and collaborate with faculty members on research projects at EPFL.
After graduating, in 2021 I joined the YCX as a Machine learning Engineer where I engaged in research and development of models and algorithms for predictive maintenance, energy generation optimization, and demand forecasting. Notably, I collaborated with two machine learning engineers and developed a deep learning-based time-to-event model for predictive maintenance in a Japanese bio-fuel power plant, which resulted in a 22% reduction in unplanned downtime. Currently, at ABC, I have furthered my skills by developing a system for automated product recommendation and marketing message generation. I utilized in-context learning with a Llama 2 model for personalized product recommendations and fine-tuned it with QLoRA for personalized marketing message generation based on product recommendations and demographic profiles. Through these professional experiences, I enhanced my skills in model development and deployment and learned how to collaborate and confidently communicate my work, preparing me for future research opportunities.
My academic and professional experiences have provided me with the knowledge, skills, and exposure required to pursue a rigorous multidisciplinary program and contribute to enhancing the reasoning capabilities of LLMs. I am confident that an education at EPFL will allow me to grow as a research engineer and contribute to research in this field.