Sam Elder
Lead Machine Learning Scientist at Kebotix- Claim this Profile
Click to upgrade to our gold package
for the full feature experience.
Topline Score
Bio
0
/5.0 / Based on 0 ratingsFilter reviews by:
Experience
-
Kebotix
-
United States
-
Chemical Manufacturing
-
1 - 100 Employee
-
Lead Machine Learning Scientist
-
Aug 2022 - Present
-
-
Machine Learning Scientist
-
Nov 2018 - Aug 2022
-
-
-
Insight Data Science
-
United States
-
Higher Education
-
1 - 100 Employee
-
Data Science Fellow
-
Sep 2018 - Nov 2018
- Produced automated analytics using Facebook Prophet for forecasting food waste to help clients of food donation startup Copia reduce their waste by up to ~ 80%. - Offered a range of forecasts aggregated weekly, monthly, and quarterly to improve predictability and match potential decision timescales for businesses. - Integrated forecasts from Python and Pandas into Periscope data visualization dashboards for use by Copia as a core component of their data analysis product. - Produced automated analytics using Facebook Prophet for forecasting food waste to help clients of food donation startup Copia reduce their waste by up to ~ 80%. - Offered a range of forecasts aggregated weekly, monthly, and quarterly to improve predictability and match potential decision timescales for businesses. - Integrated forecasts from Python and Pandas into Periscope data visualization dashboards for use by Copia as a core component of their data analysis product.
-
-
-
Massachusetts Institute of Technology
-
United States
-
Higher Education
-
700 & Above Employee
-
Graduate Student Researcher
-
Aug 2012 - Aug 2018
- Designed extrapolation methods for eliminating the bias inherent in the smaller training set sizes all classical cross-validation methods utilize, improving validation accuracy in an online learning context. - Built hierarchical cross-validation methods to select between the growing variety of cross-validation methods (new and classical) in a data-dependent fashion. - Discovered difficulties with generalization under adaptivity, producing new limits on when obfuscation (e.g. adding noise) can preserve holdout set accuracy despite repeated, adaptive use. - Proved that the ubiquitous Beta distribution is tightly concentrated about its mean. Show less
-
-
Education
-
Massachusetts Institute of Technology
Doctor of Philosophy (PhD), Applied Mathematics -
Caltech
Bachelor of Science (BS), Mathematics and Chemistry