I’m a Data Science graduate student at Stony Brook University and a Research Software Engineer at the SUNY Research Foundation, working within the Department of Psychology at Stony Brook Medicine. My work focuses on building scalable machine learning and data engineering systems for clinical and behavioral health research. I specialize in multimodal analytics, working with audio, text, mobility, and sensor data to support large-scale psychiatric studies. My interests sit at the intersection of applied machine learning, natural language processing, and real-world healthcare impact, with a strong emphasis on building robust, production-ready research pipelines.

I’m particularly interested in applied machine learning and data science problems where models intersect with real-world systems and human behavior. This includes building reliable ML pipelines for healthcare and behavioral research, especially in settings involving noisy, longitudinal, and multimodal data. I’m actively looking to learn more about advanced natural language processing and large language models, including evaluation, alignment, and interpretability of LLM-driven systems in high-s

I’m currently working as a Research Software Engineer at SUNY Research Foundation within the Department of Psychology at Stony Brook Medicine. My focus is on building scalable, production-grade data pipelines for large clinical and behavioral health studies. Right now, I’m tinkering with multimodal analytics systems that combine mobile sensor data, geospatial signals, and clinical audio. This includes designing RAPIDS-based pipelines for GPS and accelerometer data to extract mobility and activi

I am a Grad student from Stony Brook University, Majoring in Data science. Im currently working as a AI systems intern at lab called "Joint Photon sciences institute", which is in tie-up with the Brookhaven National Laboratory. Additonally im also working as research assistant at two labs in the university, namely, "LUNR" and " Stony Brook blockchain bussiness lab". Before this i did my Undergrad in India, Majoring in Automotive engineering with a Honours in AI. I had expereince working as a Computer vision and simulation engineer in a university called "Trier University of applied sciences" in germany as a part of my DAAD program in my undergrad, and i also worked as Data analyst in KPMG, before starting my Masters. I had experience working in multiple disciplane with extreme adaptation.

My current area of interest lies in the intersection of AI and systems. I had/am working on projects that deals with inference optimization, Agent orchestration, Fine tuning/continued pre-training and edge level deployment of LLMs. I love going into the nitty gritty details needed at edge computing and thurst for challenges it offers to me.

AI based ide assitants like cursor, copilot(via vscode) are awesome, but they always have this problem with token limits, which i had enough of. So to tackle that, im working on building my own local assistant that does the work cursor does, with no token limit, complete security and fully sandboxed. I call it "Sandy", similar to cursor and co-pilot, sandy is agentic, its runs on a teacher-student based paradigm. I am a solo dev working on this in my free time, so its still a work in progress.

Naman Deep is a Data Science graduate student at Stony Brook University and a Data Analyst at Stony Brook Medicine, analyzing wearable data for sleep pattern research. Previously, as a Data Scientist at Deloitte, he built data-driven solutions including lead clustering, demand forecasting, NLP for customer complaint analysis, statistical models for early warranty issue detection, and an LLM-powered supply chain assistant. At Coforge, he's focusing on a root cause analysis system for microservices using GNNs and fine-tuning LLMs with graph-based retrieval for incident resolution. He is passionate about applying AI to solve practical problems and improve decision-making and automation in organizations.

AI Evaluation

Currently working on a root cause analysis system for microservices using graph neural networks and fine-tuning lightweight language models integrated with a graph-based retrieval system at Coforge. Previously developed an LLM-powered supply chain assistant at Deloitte. Also published work on 'Deep Learning based Malignant Melanoma Detection in Dermoscopy Images'. Currently also working on AI Eval and building agentic workflows.

I’m a Master’s student in Data Science at Stony Brook University and enjoy building systems where AI meets real user workflows. My background spans software engineering, data systems, and applied AI, from payment infrastructure reliability to agentic AI pipelines and LLM applications. Lately my interest has shifted toward interaction design for intelligent systems, not just making models work, but making them understandable and usable. I like building practical tools, prototyping interfaces quickly, and turning abstract AI capability into something people can actually collaborate with.

Interested in AI interaction patterns beyond chat, interfaces shaped by model reasoning. Curious about adaptive UIs, human-AI collaborative workflows, visualizing multistep reasoning, and turning outputs into actions instead of text. Excited to explore unconventional but intuitive UI ideas and collaborate with designers/front-end builders who like experimenting with interaction models, not just features.

Right now I’m building a multi-intent AI assistant that can handle messy real conversations instead of one-prompt-one-answer flows. If a user asks something like compare options, save a few, and calculate cost in another currency in one message, the system detects multiple goals, breaks them into tasks, and keeps shared context across them. The part I’ve been most interested in is the interface how the UI should behave when the AI is juggling several intentions at once. Instead of a long chat