Sedrick Bouknight's Projects
Convolutional Neural Network trained on CIFAR10 dataset utilizing TensorFlow framework and Keras API to classify thousands of images into ten distinct categories
Built a regression model analyzing concrete strength with Tensorflow and Keras
Completed dialogue summarization task using Generative AI. Utilized FLAN-T5 Large Language Model using Hugging Face Transformers and experimented with zero-shot, one-shot, and few-shot inference. Performed PEFT fine-tuning to improve inferences and used ROUGE metrics for evaluation and reinforcement learning to generate more positive summaries.
Utilizes Reinforcement Learning, specifically Deep-Q Learning, and OpenAI Gym to train an agent to perform well in the popular first-person shooter game 'Doom'
A series of programs coded in Python using OpenCV, Tensorflow backend, and open source neural networks to complete tasks including facial location, expression classification, age/gender estimation, and real-time facial recognition.
Utilizes Hough Theory for lane detection in self-driving car application
Used Keras to solve a variety of problems using real-world datasets. Covered classification tasks including disease detection and truck failure and regression tasks such as toxicity prediction. Built CNN models for image classification and used RNNS for stock price prediction.
From past Kaggle problem. Predicts the effect of Genetic Variants to enable Personalized Medicine using various machine learning and natural language processing techniques.
Hands-on workshop with various Machine Learning algorithms, tools, and techniques including datapreprocessing, supervised/unsupervised learning, artificial neural networks, and model production.
Neural Network trained on MNIST dataset to recognize handwritten digits
An organic chemistry chatbot made using OpenAI's Whisper API and Eleven Labs' TTS API. Developed with React and FastAPI.
Utilizes OpenCV functions to display various edge detection methods inlcuding Sobel, Laplacian, and Canny edge detection in real time via webcam
Transfer learning using pretrained ResNet18 architecture on PyTorch framework to classify images for concrete crack detection
Used Keras and TensorflowHub for various transfer learning tasks including image and text classification with Google Colab