Ahsan Ashraf
Data Scientist + Physicist
About
I am a data scientist and physicist, currently working on trust and content integrity at Airbnb. I am also an assistant professor of the practice of data science at the data science institute at Brown University. I develop and apply machine learning and statistical methodologies/tools to tackle modern data science problems, build data products, and enable data-driven/informed decision making.
Previously I led a team of data scientists working on homefeed ranking / recommendations and search quality at Pinterest, and I was an adjunct faculty at UC Berkeley's school of information. Before that I worked with a personal finance start-up called wallet.ai as part of a data science fellowship at Insight. My research background is in complex systems and condensed matter physics. I worked at Brookhaven National Lab on the development of novel sustainable energy technologies. More details on my PhD research can be found here.
Highlights of recent work
TWiML&AI Podcast: Diversification in Recommender Systems
In our conversation, Sam and I discuss experiments we ran to explore the impact of diversification in homefeed, the methodology my team used to incorporate variety into the Pinterest recommendation systems, the metrics we monitored through the process, and the business impact of the work.
Lead the effort to develop and deploy machine learning algorithms to detect and remove harmful content (self-injury, medical misinformation, and hate speech among others) from major discovery surfaces
Developed comprehensive health metrics for topical diversity - in addition to relevance - in the home feed to prevent over-optimization of ranking/recommendation algorithms to short-term outcomes at the expense of long-term product health (exploration versus exploitation), helping drive the understanding and future strategy of personalization for machine learning teams at Pinterest
Built a recommender system for early stage startup wallet.AI, as part of data science fellowship at Insight, that builds intelligent machines to help people make better financial decisions
Developed a new architecture for electronic devices for graphene-semiconductor junctions using n-doped multi-layer graphene on glass
Designed and built a new method for texturing a surface at the nanoscale (< 50nm) using self-assembly of polymers for antireflection in silicon (and other) solar cells at any angle or wavelength of light
Developed a novel way of mapping where the charges in electronic devices are being generated and collected to better understand how to optimize organic photovoltaic devices
Press Coverage:
