I am a technology professional with an academic research and industry background in Natural Language Processing, Deep Learning, Machine Learning, Differential Privacy and Software Development.
I currently work as a Machine Learning Engineer on the research team of Pivotus Ventures, a Fintech innovation lab. My responsibilities include (but are not limited to) reproducing results from research papers, conceptualizing and building Proof-Of-Concept AI solutions to business problems and conducting blue-sky research. Major accomplishments in the current role include building an abusive content flagging system using deep convolutional networks that bettered the state-of-the-art.
I have graduated from SUNY-University at Buffalo with a Masters Degree in Computer Science. My thesis investigated the use of differential privacy to tackle overfitting in deep neural networks.
In the past, I have worked as a Data Science Intern to build a neural network based regression / classification system and a 'model-as-configuration' based deployment system for neural networks. Over the summer of 2016, I interned at a platform-based solar energy company in Oakland,CA. My role was primarily DevOps with additional responsibilities extending into the domains of Site Reliability and Build & Release engineering.
I have completed academic coursework in the areas of Machine Learning, Differential Privacy, Multi-lingual Information Processing and Retrieval, Pattern Recognition, Deep Learning, Differential Privacy, Computer Vision and Networks. I have worked as a student researcher to develop a customized search engine for Arabic Script data.
During my time off, I enjoy travelling, mountaineering, landscape photography, skiing and philately.
Machine Learning Engineer (Research)
What are you doing to control abusive content on your platform? Can your current solution tell the difference between “f**king awesome” and “f**king loser”? Can it detect racist and sexist content?
You will learn how to build a deep learning based solution and deploy it as a micro-service.
This talk focuses on :
The shortcomings / common pitfalls of current approaches in terms of:
- Handling the ambiguities in defining offensive content
- Using a Deep Learning based approach to detect offensive content
- Production-izing the solution