Pawniwillaceit
2 days ago
Graduate / Computational modelling and Human-Computer Interaction - International Research Programs [2]
Hi all, I'm currently in my 3rd year of a Bachelor's degree and have been actively applying for an international research internship. I feel as though I'm still lacking when it comes to SOP and research statements. Would greatly appreciate any feedback from the community!
Research Statement
I am currently in my 3rd year of B.Tech in Information Technology. My research interest lies at the intersection of Computational modelling and Human-Computer Interaction (HCI), with an aim to design LLM‑powered intelligent interfaces that can more naturally blend our physical and digital worlds.
As an undergraduate research assistant in the SCALE Lab at the Indian Institute of Science, I collaborated on a project to improve the LLM's performance as well as its understanding of user queries or prompts. This experience led me to inspect both model-centric improvements and the user-centric aspects of query interpretation, as well as the deeply embedded connection of LLM output with the quality of user input.
Additionally, my research experience at QWorld as a Quantum Machine Learning Intern and a cybersecurity research intern at DRDO has made me deeply interested in how intelligent systems can model and aid humans in complex, uncertain environments.
Equipping me with a strong background in subdomains of Artificial Intelligence like machine learning, Reinforcement Learning, neural networks, transformers, Large Language models, etc, these experiences have assisted me in developing an interest in interactive AI and LLM‑based systems that better understand human context, and behaviour so that they can support more robust decision‑making.
Human-LLM interaction, in the context of conversational LLMs, has been limited even after the proliferation of advanced technology and intelligent models. This is due to ineffective inputs which either fail to provide enough context or are not in legible English, which is a challenge for people who do not have a thorough proficiency in English. The fact remains that we know how to operate an LLM, not how to converse with it. Large language models have enabled fluent conversational systems, but aligning their behaviour with conversational goals, user preferences, and emotional context remains a challenge.
To draw an analogy, it is like talking to an all-knowing toddler, although she may have all the knowledge in the world, if you don't explain your query coherently and intelligibly, along with full context, the query answer would be inefficient and inconsistent with user expectations.
I believe with the rise of conversational LLMs, there is an urgent need to develop and advance HCI. With the base idea of developing a framework to assist users in asking the right questions, I collaborated in the development of Promptqueen, which is a multi-agentic intelligent interpreter capable of converting Rudimentary, Ambiguous, and Weak (RAW) natural language into Structured, Organized, and Detailed (SAD) prompts to optimise output optimality.
This was supported by a question loop, which clarified basic knowledge gaps that LLM had for a given user query and a feedback loop, which ensured quality assurance for a generated user prompt. To deal with Issues like verbosity and linguistic diversity, we developed a markup language, POML (Prompt Orchestration Markup Language), following the release of Microsoft's POML research paper. Human personality modelling via life graph and emotion tracking proved to be of use in increasing output relevance for a user query.
During the {internship_name} internship, I would like to build on this base by exploring how reinforcement learning, human feedback, and behavioural modelling can be combined with LLMs to create autonomous agents that adapt to humans.
In particular, I am interested in studying how interactive feedback, preferences, and models with an understanding of human personality can shape conversations with AI, and in investigating how these ideas can be applied to domains such as robot autonomy, autonomous vehicles, and dialogue systems
{after this i include how different professors work align with my interests}
Following my research interest, I am deeply interested in the works of {name} on integrating human feedback into robot learning and decision‑making closely matches my interest in human‑aided reinforcement learning and interactive autonomy. My experience with RL‑based simulation, multi‑agent AI pipelines, and modelling user input would allow me to contribute to projects that study how robots learn from the preferences of a human.
Hi all, I'm currently in my 3rd year of a Bachelor's degree and have been actively applying for an international research internship. I feel as though I'm still lacking when it comes to SOP and research statements. Would greatly appreciate any feedback from the community!
Research Statement
I am currently in my 3rd year of B.Tech in Information Technology. My research interest lies at the intersection of Computational modelling and Human-Computer Interaction (HCI), with an aim to design LLM‑powered intelligent interfaces that can more naturally blend our physical and digital worlds.
As an undergraduate research assistant in the SCALE Lab at the Indian Institute of Science, I collaborated on a project to improve the LLM's performance as well as its understanding of user queries or prompts. This experience led me to inspect both model-centric improvements and the user-centric aspects of query interpretation, as well as the deeply embedded connection of LLM output with the quality of user input.
Additionally, my research experience at QWorld as a Quantum Machine Learning Intern and a cybersecurity research intern at DRDO has made me deeply interested in how intelligent systems can model and aid humans in complex, uncertain environments.
Equipping me with a strong background in subdomains of Artificial Intelligence like machine learning, Reinforcement Learning, neural networks, transformers, Large Language models, etc, these experiences have assisted me in developing an interest in interactive AI and LLM‑based systems that better understand human context, and behaviour so that they can support more robust decision‑making.
Human-LLM interaction, in the context of conversational LLMs, has been limited even after the proliferation of advanced technology and intelligent models. This is due to ineffective inputs which either fail to provide enough context or are not in legible English, which is a challenge for people who do not have a thorough proficiency in English. The fact remains that we know how to operate an LLM, not how to converse with it. Large language models have enabled fluent conversational systems, but aligning their behaviour with conversational goals, user preferences, and emotional context remains a challenge.
To draw an analogy, it is like talking to an all-knowing toddler, although she may have all the knowledge in the world, if you don't explain your query coherently and intelligibly, along with full context, the query answer would be inefficient and inconsistent with user expectations.
I believe with the rise of conversational LLMs, there is an urgent need to develop and advance HCI. With the base idea of developing a framework to assist users in asking the right questions, I collaborated in the development of Promptqueen, which is a multi-agentic intelligent interpreter capable of converting Rudimentary, Ambiguous, and Weak (RAW) natural language into Structured, Organized, and Detailed (SAD) prompts to optimise output optimality.
This was supported by a question loop, which clarified basic knowledge gaps that LLM had for a given user query and a feedback loop, which ensured quality assurance for a generated user prompt. To deal with Issues like verbosity and linguistic diversity, we developed a markup language, POML (Prompt Orchestration Markup Language), following the release of Microsoft's POML research paper. Human personality modelling via life graph and emotion tracking proved to be of use in increasing output relevance for a user query.
During the {internship_name} internship, I would like to build on this base by exploring how reinforcement learning, human feedback, and behavioural modelling can be combined with LLMs to create autonomous agents that adapt to humans.
In particular, I am interested in studying how interactive feedback, preferences, and models with an understanding of human personality can shape conversations with AI, and in investigating how these ideas can be applied to domains such as robot autonomy, autonomous vehicles, and dialogue systems
{after this i include how different professors work align with my interests}
Following my research interest, I am deeply interested in the works of {name} on integrating human feedback into robot learning and decision‑making closely matches my interest in human‑aided reinforcement learning and interactive autonomy. My experience with RL‑based simulation, multi‑agent AI pipelines, and modelling user input would allow me to contribute to projects that study how robots learn from the preferences of a human.
