diff --git a/README-imichel.md b/README-imichel.md new file mode 100644 index 0000000..f58036d --- /dev/null +++ b/README-imichel.md @@ -0,0 +1,37 @@ +## **Introduction** +Hello! My name is Isabel and I am a CS Master's student specializing in AI. As a first-generation low-income student in CS with a non-traditional background, I am looking forward to committting to this course and going far with my assigned project! Below are my technical skills and projects/experiences. + +## **Technical Skills** +- **Languages & Tools:** Python, C++, Java, PyTorch, Google Cloud Platform, AWS, Linux, Git. +- **Technical Skills:** Deep Learning for Computer Vision, Spoken & Natural Language Processing, Decision Making Under Uncertainty, Data Mining & Analysis. + +--- + +## **Projects** +### **Cortex-Level Brain MRI Generation Using Diffusion Models** +*Spring 2024* +- Collaborated with a mentor and team at Stanford's Computational Neuroscience Lab to develop a two-stage ML pipeline for generating high-fidelity synthetic brain MRIs. +- Preprocessed 3060 MRIs through denoising, bias correction, skull stripping, affine registration, normalization, and data augmentation. +- Employed unsupervised deformable image registration with diffusion models to learn deformation fields, achieving strong performance relative to baseline models. +- Project report publicly available [here](https://cs231n.stanford.edu/2024/papers/cortex-level-brain-mri-generation-using-diffusion-models.pdf). + +### **Cercano: A Parking Spot Near You** +- Developed a decision-making system leveraging reinforcement learning techniques to recommend the closest parking spot. The decision-making policy outperformed random-action by 50% in major urban centers. +- Refined the policy based on simulated parking dynamics, integrating 34,000+ parking structures from public data. +- Used random sampling techniques to position cars within geographic boundaries and optimized parking recommendations based on driving distance and time retrieved via the Google Maps API. +- Project report publicly available [here](https://aa228.stanford.edu/old-projects/). + +### **Humor Me!: Detecting Humor in Stand-up Comedy** +- Built a humor detection tool for binary classification on audio input using fine-tuned foundational models (Wav2Vec 2.0, Hubert, Conformer). +- Analyzed ~40 hours of multi-modal, humor-annotated data. Addressed class imbalance and evaluated robustness in misclassification cases. +- Conducted comparative analysis between audio-based and text-based classifiers (DistilBERT). + +### **Automated Hate Speech Detection Bot** +- Developed a Discord bot targeting hate speech against women using OpenAI's GPT-4 API and prompt engineering for improved accuracy. +- Presented the project to 15+ industry representatives, showcasing real-time content evaluation and automated actions to enhance user safety. + +--- +## **Research Experience** +### **Research Assistant** +*Esfahani Lab at Stanford University* +- Conducting research on early cancer detection using computer vision to identify potential cancer biomarkers.