Machine Learning Projects

ML & Data Science Portfolio

I’m passionate about building predictive models and performing causal inference in real-world, often noisy and biased datasets. My expertise lies in probabilistic modeling, particularly through generative models and Bayesian methods. Below are some sample projects I led end-to-end — from data processing to model development, evaluation, and deployment. Feel free to reach out if you’d like to chat: sumqezlou+at+gmail.com

Python TensorFlow PyTorch Scikit-learn Keras NumPy Pandas HPC/GPU Docker Git

Accurate Causal Inference using Surrogate Modeling

GenAI for Surrogate Modeling

Gaussian Processes Deep Learning Variational Inference TensorFlow Distributed Training-GPU/CPU Uncertainty Quantification

Developed two independent surrogate models (Gaussian Processes and Deep Learning) to map cosmological simulation outputs to realistic observed data. These novel models enable fast, percent-level accurate inference on observational datasets, for the first time.

Gaussian Processes Model Deep Learning Model Interactive Demo

Robust GenAI for Signal Detection in Noisy, Biased Data

Autoencoders Time Series Bayesian Model Selection PyTorch Generative Models

Built a generative probabilistic model to separate subtle signals from systematics and noise, enabling reliable detection in noisy data.

View Code Tutorials

Image Recognition in 3D Noisy Map Denoising

Autoencoders TensorFlow 3D CNNs Signal Processing

Built a 3D convolutional autoencoder and classic denoising algorithms to denoise large-scale tomographic reconstructions. Achieves 40% noise reduction while preserving structural features.

View Code

Bayesian Inference & Probabilistic Models

Bayesian ML MCMC Uncertainty Quantification Python

Implemented Bayesian inference pipelines for parameter estimation with uncertainty quantification. Uses MCMC sampling (emcee, PyMC3) and variational inference. Applied to cosmological parameter constraints and model selection.

View Code