shamikbose / mlfromscratch

A repository of Machine Learning algorithms from scratch

Hi there 👋

I'm Shamik and I enjoy building solutions to problems, mostly through programming (and occasionally with WD-40). I work as a Lead Data Scientist building machine learning applications for detecting and anonymizing PII and PHI in data breaches. I am also a part-time contributor to the BigScience Workshop, the BigBIO effort and the BigCode Project from 🤗. In addition, I am working with PIISA, a collection of data scientists, software developers and lawyers to establish an open standard for PII protection that can be used across the globe. You can follow our efforts here. I also like to cook 👨‍🍳

├── Interests
│   ├── Natural Language Processing
│   ├── Explainable Machine Learning
│   ├── AI Ethics
│   ├── System Design
│   └── PII Anonymization
├── Occupations
│   ├── Software Engineer
│   ├── Graduate Research Assistant
│   ├── Lead Data Scientist
│   └── Senior Researcher
├── Locations
│   ├── Kolkata, India
│   ├── Boston, MA, USA
│   ├── Tallahassee, FL, USA
│   └── Leeds, England
└── Book Suggestions
    ├── Fiction
    │   ├── The Three Body Problem - Cixin Liu
    │   ├── All the Light we cannot see - Anthony Doerr
    │   └── Purple Hibiscus - Chimamanda Ngozi Adichie
    ├── Non-Fiction
    │   ├── Algorithms of Oppression - Safiya Umoji Noble
    │   ├── Braiding Sweetgrass - Robin Wall Kimmerer
    |   ├── Chaos Machine - Max Fisher
    |   ├── Viral Justice - Ruha Benjamin
    │   └── Weapons of Math Destruction - Cathy O. Neill
    └── Cookbooks
        ├── The Food Lab - J. Kenji Lopez-Alt
        ├── Mi Cocina - Rick Martinez
        └── Dessert Person - Claire Saffitz

Projects

1. Scientific Title Generator
2. BigBIO dataloaders
3. MIT 6.006 Solution Notebooks

Publications

1. Explaining AI for Malware Detection: Analysis of Mechanisms of MalConv
2. PhD Thesis: Towards Explainability in Machine Learning for Malware Detection
3. Static Malware Modeling and Detection using Topic Models
4. BigBIO: A Framework for Data-Centric Biomedical Natural Language Processing
5. The bigscience roots corpus: A 1.6 tb composite multilingual dataset

P.S. The tree was built using Rich