I'm Mrunal, an ML Engineer and AI grad student at Northeastern University. I love building ML systems, training models and working with messy data.
Website / LinkedIn / Kaggle / Email
class Mrunal:
role = "ML/AI Engineer, MS in AI @ Northeastern University"
focus = ["core ML systems", "semantic layers over messy data", "NLP", "RL" ]
previously = "AI/ML Intern @ Silver Touch: NLP for legal compliance"
research = "Pneumonia detection from chest X-rays"
datasets = "2 published on Kaggle, 500+ downloads"
building = "Geosfera"
def talk_about(self):
return ["Bits", "Atoms", "Neurons", "Genes"] # anything BANGAI from First Principles
ML algorithms built from scratch, one notebook at a time.
[notes]
[libcode]
linear regression ✓
Detecting Pneumonia with Dual-Path Ensembled Deep Learning Models with Lung Segmentation on Chest X-ray Images
U-Net lung segmentation feeding an ensemble of CNN classifiers.
[paper]
[segmentation code]
[detection code]
[CRD dataset]
[PneumoniaCXR dataset]
Geosfera
Android geography game for learning countries, capitals, and flags.
Artificial Intelligence and Data
- AI from First Principles: ML algorithms from scratch, with notes and a modular library
- Pneumonia detection research: dual-path ensembled CNNs with U-Net lung segmentation on chest X-rays, with code for segmentation and detection and two datasets, CRD and PneumoniaCXR
Mobile
- Geosfera: Android geography game for learning countries, capitals, and flags
- Chronicles: AI-powered diary app with speech-to-text, text-to-speech, to-dos, and story sharing