Deekshitth Vegi
An AI developer and researcher building intelligent systems that bridge human creativity and machine logic.
From deep learning architectures to full-stack applications, I turn complex data streams into tangible, interactive experiences. My long-term goal is to architect a fully autonomous 3D AI that learns, adapts, and creates.
Experience & Research
Education
University of North Texas – M.S. in Artificial Intelligence
Aug 2023 – Jul 2025
Thesis: Automated Methods to Evaluate the Quality of Distractors in College-Level MCQs using LLMs
IIITDM Kurnool – B.Tech. in Mechanical Engineering
Aug 2019 – May 2023
Research
Graduate Thesis Researcher — UNT (Aug 2024 – Jul 2025)
- Developed automated methods to evaluate distractor quality in MCQs.
- Compared LLM-as-Judge vs LLM-as-Examinee paradigms, benchmarking models like GPT-4, Claude, and LLaMA-2.
- Found Claude 3.5 Sonnet optimal for human alignment and deployed the system on Hugging Face for public use.
Work Experience
Data Science Intern — Innomatics Research Labs
Sep 2022 – Jun 2023
Built NLP pipelines for tweet classification and designed a Spotify recommender system using cosine similarity.
Software Developer Intern — Josh Innovations
Jun 2022 – Oct 2022
Developed full-stack applications using ReactJS & Node.js, implementing real-time UI interactions.
Research Intern — DRDO, Ministry of Defense
May 2022 – Oct 2022
Designed and analyzed composite security structures, delivering prototypes meeting defense standards.
Tech Stack
Programming
Web Dev
AI/ML
Data & Tools
Core Projects
Exoplanet Transit Detector
LSTM/Transformer classification on a dataset of 150K light curves, with a 3D orbit simulation UI built with Gradio.
2D → 3D Pose Estimation
Extracted 2D keypoints from images to reconstruct 3D human poses using advanced filtering and regression models.
Wikipedia Image Captioning
Compared BLIP/CLIP embeddings with a custom LSTM decoder for caption generation, deployed on Hugging Face.
Chat with AI Deekshitth
Key Achievements
Deployed 4 AI models on Hugging Face Spaces.
Built a 3D visualizer for exoplanet transits.
Led a 500+ question LLM benchmark study with statistical validation.
Implemented foundational deep learning components using Minitorch.
Personal Interests
🎮
Gaming
Competitive titles, especially Valorant.
🍿
Anime & Movies
One Piece 🏴☠️ & Interstellar.
🌍
Travel
Favorite Destination: Japan.
🎵
Music
The Weeknd. Attended his concert in Dallas on 28th Aug 2025. First time drunk, a peak life moment.