Nikhil
Marisetty

I am a CS student at UT Dallas excited about exploring innovations in AI and ML, and I strive to consistently grow both technically and personally through new challenges.

4.0

Grade Point Average

Hackathons Won

20+

Projects Done

NIKHIL MARISETTY

Work Experience

Jun 2025 - Present

CoachVision.ai

Founder & CEO

Building an AI-powered coaching platform for sports that analyzes and provides personalized feedback on form and technique.

Jun 2025 - Present

CoachVision.ai

Founder & CEO

Building an AI-powered coaching platform for sports that analyzes and provides personalized feedback on form and technique.

Feb 2026 - Present

UT Dallas

Soft Robotics Research Assistant

Working on building object detection and pick and place systems for flexible, life-like machines which use smart materials to mimic biological muscles

Feb 2026 - Present

UT Dallas

Soft Robotics Research Assistant

Working on building object detection and pick and place systems for flexible, life-like machines which use smart materials to mimic biological muscles

Aug 2025 - Present

UT Dallas

Neuromorphic AI Hardware Research Assistant

Working on building an architecture where brain-inspired chips process information in pulses and spikes in contrast to backpropagation.

Aug 2025 - Present

UT Dallas

Neuromorphic AI Hardware Research Assistant

Working on building an architecture where brain-inspired chips process information in pulses and spikes in contrast to backpropagation.

Sept 2025 - Present

UT Dallas

CS Grader & TA

Provide grading and instructional support for the Systems Programming course, helping students strengthen their understanding of system-level programming.

Sept 2025 - Present

UT Dallas

CS Grader & TA

Provide grading and instructional support for the Systems Programming course, helping students strengthen their understanding of system-level programming.

Jun 2025 - Aug 2025

NRVE

AI Engineer Intern

Integrated an AI music generation model, DiffRythm, and built a personalized chatbot using the Gemini API for users explore music production.

Jun 2025 - Aug 2025

NRVE

AI Engineer Intern

Integrated an AI music generation model, DiffRythm, and built a personalized chatbot using the Gemini API for users explore music production.

Jan 2025 - May 2025

Cohort Science

Software Engineering Intern

Led a group of interns in developing a data-driven platform that collects and analyzes college information to generate insights to improve enrollment, retention, and strategic planning.

Jan 2025 - May 2025

Cohort Science

Software Engineering Intern

Led a group of interns in developing a data-driven platform that collects and analyzes college information to generate insights to improve enrollment, retention, and strategic planning.

Showcasing My Best Work

From AI-powered solutions to ML model deployment, each piece highlights my dedication to crafting intelligent, functional, and innovative solutions tailored to client needs.

CoachVision.ai

An AI sports recommendation platform that analyzes user form and technique (across 8 different sports) using Computer Vision libraries and frameworks such as OpenCV and Mediapipe to provide feedback on the strengths and areas of improvement for the user.

AutoPM

End-to-end product management workflow that leverages LangGraph agents and the Gemini API to orchestrate specialized agents to handle everything from research-backed idea generation and market analysis to the creation of detailed user stories and interactive wireframes. By integrating automated RICE prioritization, OKR alignment, and seamless Jira ticket generation, the workflow ensures that strategic goals are instantly translated into actionable, well-documented technical tasks.

SupplyNet

SupplyNet is a platform that turns complex economic and geopolitical data into actionable insights. Engineered 50+ features across unemployment, markets, and macro trends. Uses PCA to create a single GSSI index from noisy datasets. XGBoost model forecasts index movements like an economic weather report and an NLP pipeline analyzes geopolitical news and convert sentiment into portfolio signals.

CorpNet

Python-based implementation of ArchRAG that constructs a sophisticated Graph-based RAG system through a two-phase architecture. During offline indexing, the system transforms a text corpus into a Knowledge Graph, utilizing the Leiden algorithm to iteratively cluster entities into a summarized community hierarchy, indexed via C-HNSW for high-speed traversal. In the online retrieval phase, CorpNet performs a top-down search of this hierarchical index, employing an LLM to adaptively filter, score, and synthesize candidate results into a contextually rich final response.

Pondr

An adaptive learning platform featuring a LangGraph agentic workflow powered by 𝗚𝗲𝗺𝗶𝗻𝗶, Tavily, and the YouTube API to map concepts into a living knowledge graph. The experience is fully voice-integrated via ElevenLabs, while an XGBoost regressor personalizes the Ebbinghaus forgetting curve to proactively schedule reviews before you forget.

TunaDance

Generative AI model that utilizes a Gaussian Diffusion framework with an 8-layer Transformer decoder to synthesize synchronized 3D full-body choreography. By processing 35-dimensional audio features via librosa, it generates high-fidelity SMPL-X motion sequences at 30 FPS, covering 52 joints including detailed hand movement. To ensure physical realism, the model was trained on the FineDance dataset using a specialized loss function

EpicRAG

AI system using a multimodal RAG framework to extract insights from complex documents, processing text, tables, and images with embeddings in ChromaDB for context-aware retrieval and response generation. Built with Python, LangChain, Streamlit, and OpenAI APIs.

Career Agents

Multi-agent AI platform hosted on AWS Bedrock AgentCore that helps UTD students plan their careers through intelligent coordination of specialized agents. These agents analyze live job market trends, and UTD course catalogs to create personalized learning and career paths.

3LC

Used 3LC to label a large, undefined dataset for a binary image-classification model. Starting with 200 labeled and 4,000+ unlabeled images, I built an iterative workflow using clustering and model-driven labeling. This process cleaned the dataset, reduced manual work, and improved model accuracy to 96.37%.

HootHive

AI-powered learning assistant that transforms recorded lectures into structured summaries, interactive Q&A chats, and auto-generated flashcards and quizzes, enabling faster review and collaborative study. Built using TwelveLabs and Gemini API.

Matcha

A full-stack AI-powered hackathon teammate matching platform. It uses a dual-backend architecture (Flask + Node.js) with Selenium and BeautifulSoup for dynamic Devpost scraping, MongoDB for persistence, JWT-based auth, and Google Gemini for intelligent teammate matching and idea generation.

MedVisit

AI-powered healthcare platform that automates the consultation lifecycle, transforming video encounters into structured, actionable data for both providers and patients. By utilizing a high-performance tech stack featuring FastAPI and Next.js, the system employs a multimodal AI pipeline—driven by TwelveLabs and ElevenLabs—to generate automated SOAP notes and EMR records for doctors while providing patients with personalized, easy-to-navigate care plans

CabinIQ

Real-time passenger stress monitoring platform that uses camera-based biometric sensing and AI to help flight crews proactively manage passenger wellbeing. It captures heart rate and breathing data via the Presage mobile app and displays it on a live dashboard with color-coded 2D and 3D seat maps, individual vitals, and stress trends across flight phases.

Fiyota

Data Visualization platform designed to empower car buyers with clarity and confidence. With Fiyota, users can compare financing vs. leasing options side-by-side, run monthly payment simulations, locate nearby Toyota dealerships, and begin pre-approval checks directly in the dashboard

FoodVision

Developed and trained a computer vision model using Convolutional Neural Networks (CNNs) for food recognition, achieving 85% top-1 accuracy and 93% top-5 accuracy on the Food-101 dataset. Designed a deep learning model capable of identifying diverse food categories with high precision.

StyleSense

Developed a multi-class classification model for fashion recognition that classifies clothing and outfit images using the Fashion-MNIST dataset. Designed and trained a CNN to accurately distinguish between apparel types such as shirts, dresses, and sneakers.

Friction

Developed a 2D obstacle course game using the Unity Game Engine and C#. Players control a sprite and go through levels filled with lava and ice.

My Skills

My technical expertise spans across cutting-edge AI/ML technologies, enabling me to deliver innovative solutions that drive business growth and digital transformation.