Niket
Gupta
AI/ML Engineer & Full-Stack Developer building intelligent systems that scale — from model training to production deployment.
Building at the intersection of AI and engineering
I'm a Computer Science graduate from the University of Houston focused on AI/ML engineering — building deep learning systems and taking them from research to working software.
My project work spans computer vision and CNNs (a dual-pipeline lung cancer detector that placed 3rd at UH's HPE Data Science Institute showcase), NLP and transformer models (multilingual fraud detection with DistilBERT), and relational data systems. I work primarily in TensorFlow, PyTorch, and Python on AWS.
I'm NVIDIA-certified in Deep Learning Fundamentals. Outside of coursework I teach — I worked as a coding instructor for kids — and run a personal home lab for systems and AI experimentation.
Coding Instructor for Children
- ▸Taught coding concepts and programming fundamentals to students ages 5–12 using Scratch, Roblox Studio, and Minecraft mods
- ▸Developed lesson plans and curriculum aligned with state technology education standards for a classroom of up to 20 students
- ▸Communicated complex technical concepts to non-technical audiences, building strong presentation and mentorship skills
- ▸Guided students through project-based learning, fostering problem-solving skills and computational thinking from an early age
IT Intern
- ▸Assisted with troubleshooting and resolved 10+ technical issues across computer systems and network infrastructure
- ▸Supported IT operations including system maintenance, hardware diagnostics, and end-user support
- ▸Gained hands-on experience with enterprise IT workflows, ticketing systems, and client-facing technical communication
Student Volunteer & Leadership
- ▸Led a team of 15+ cadets in community service and leadership development activities
- ▸Developed strong communication, organization, and team coordination skills
AI-Powered Smart Mirror
A smart mirror that uses MediaPipe for hand tracking and pose detection, with Kinect depth sensing for person recognition and auto-wake. I'm architecting a multi-GPU pipeline (GTX 1070 + GTX 1060) that dedicates separate cards to real-time computer vision and display rendering, with planned photo-to-3D garment conversion and weather-aware outfit recommendations running on local AI inference.
Dual-Pipeline Lung Cancer Detection System
TensorFlow CNN system that detects and classifies lung cancer across 15 distinct classes from a ~5K–15K-sample Kaggle medical imaging dataset. A dual-pipeline architecture fuses CT-scan images with clinical tabular data into a combined prediction layer, reaching ~81% validation accuracy and outperforming single-pipeline baselines.
Multilingual Fraud Detection Analysis
First systematic analysis of language diversity in the DIFrauD fraud dataset. Benchmarked DistilBERT, Random Forest, and LinearSVC on multilingual fraud data and built a langdetect + spaCy cross-validation framework with 99.77% agreement, using chi-square tests and machine-translated datasets for rigorous hypothesis testing.
Zoo Management Database System
Comprehensive MySQL database for zoo operations — managing staff, gift-shop inventory, and a ticket-booking system. Built on a normalized 3NF schema with efficient relational design and database triggers that automate business logic and preserve data integrity.