Hani Ramezani
Lead MLE/DS at Earnest- Claim this Profile
Click to upgrade to our gold package
for the full feature experience.
Topline Score
Bio
Shane (John) L.
Hani joined my team as the first data scientist at Zume, brining with him a wealth of knowledge about simulation, operations research, and machine learning. His first order of business was to get up to speed on demand forecasting; within a few weeks he was an expert and was working on state of the art models. A few weeks later he had learned the fundamentals of production deployments. Within a few months he was solid in deploying production models. I asked him to start mentoring our interns and within a few months both were performing to the point that they were offered full time jobs. This describes Hani in a nutshell: he starts strong, and keeps getting stronger. He's independent, team oriented, and awesome to work with; and if you get him on your team and give him a tough problem, you can count on him to continue plugging away at whatever gaps exist until you have a high performing solution. I'd go out of my way to work with him again, and strongly recommend him to anyone interested in a top notch scientist.
Paul Elliott
As a data scientist, Hani is a consummate professional. I worked with Hani at Zume for nearly a year, where we collaborated on machine learning solutions. As the data science owner of our demand prediction services, Hani balanced ceaseless customer focus with a keen roadmap and led us to success in every market. Both knowledgeable and direct, Hani is adept at devising the right strategy and clearly communicating to both team and stakeholders. He does all of this with a sense of humor and friendly demeanor, making him a real pleasure to work with. Hani has been an exceptional team member, co-worker and leader, and any data science team would be lucky to have him aboard.
Shane (John) L.
Hani joined my team as the first data scientist at Zume, brining with him a wealth of knowledge about simulation, operations research, and machine learning. His first order of business was to get up to speed on demand forecasting; within a few weeks he was an expert and was working on state of the art models. A few weeks later he had learned the fundamentals of production deployments. Within a few months he was solid in deploying production models. I asked him to start mentoring our interns and within a few months both were performing to the point that they were offered full time jobs. This describes Hani in a nutshell: he starts strong, and keeps getting stronger. He's independent, team oriented, and awesome to work with; and if you get him on your team and give him a tough problem, you can count on him to continue plugging away at whatever gaps exist until you have a high performing solution. I'd go out of my way to work with him again, and strongly recommend him to anyone interested in a top notch scientist.
Paul Elliott
As a data scientist, Hani is a consummate professional. I worked with Hani at Zume for nearly a year, where we collaborated on machine learning solutions. As the data science owner of our demand prediction services, Hani balanced ceaseless customer focus with a keen roadmap and led us to success in every market. Both knowledgeable and direct, Hani is adept at devising the right strategy and clearly communicating to both team and stakeholders. He does all of this with a sense of humor and friendly demeanor, making him a real pleasure to work with. Hani has been an exceptional team member, co-worker and leader, and any data science team would be lucky to have him aboard.
Shane (John) L.
Hani joined my team as the first data scientist at Zume, brining with him a wealth of knowledge about simulation, operations research, and machine learning. His first order of business was to get up to speed on demand forecasting; within a few weeks he was an expert and was working on state of the art models. A few weeks later he had learned the fundamentals of production deployments. Within a few months he was solid in deploying production models. I asked him to start mentoring our interns and within a few months both were performing to the point that they were offered full time jobs. This describes Hani in a nutshell: he starts strong, and keeps getting stronger. He's independent, team oriented, and awesome to work with; and if you get him on your team and give him a tough problem, you can count on him to continue plugging away at whatever gaps exist until you have a high performing solution. I'd go out of my way to work with him again, and strongly recommend him to anyone interested in a top notch scientist.
Paul Elliott
As a data scientist, Hani is a consummate professional. I worked with Hani at Zume for nearly a year, where we collaborated on machine learning solutions. As the data science owner of our demand prediction services, Hani balanced ceaseless customer focus with a keen roadmap and led us to success in every market. Both knowledgeable and direct, Hani is adept at devising the right strategy and clearly communicating to both team and stakeholders. He does all of this with a sense of humor and friendly demeanor, making him a real pleasure to work with. Hani has been an exceptional team member, co-worker and leader, and any data science team would be lucky to have him aboard.
Shane (John) L.
Hani joined my team as the first data scientist at Zume, brining with him a wealth of knowledge about simulation, operations research, and machine learning. His first order of business was to get up to speed on demand forecasting; within a few weeks he was an expert and was working on state of the art models. A few weeks later he had learned the fundamentals of production deployments. Within a few months he was solid in deploying production models. I asked him to start mentoring our interns and within a few months both were performing to the point that they were offered full time jobs. This describes Hani in a nutshell: he starts strong, and keeps getting stronger. He's independent, team oriented, and awesome to work with; and if you get him on your team and give him a tough problem, you can count on him to continue plugging away at whatever gaps exist until you have a high performing solution. I'd go out of my way to work with him again, and strongly recommend him to anyone interested in a top notch scientist.
Paul Elliott
As a data scientist, Hani is a consummate professional. I worked with Hani at Zume for nearly a year, where we collaborated on machine learning solutions. As the data science owner of our demand prediction services, Hani balanced ceaseless customer focus with a keen roadmap and led us to success in every market. Both knowledgeable and direct, Hani is adept at devising the right strategy and clearly communicating to both team and stakeholders. He does all of this with a sense of humor and friendly demeanor, making him a real pleasure to work with. Hani has been an exceptional team member, co-worker and leader, and any data science team would be lucky to have him aboard.
Experience
-
Earnest
-
United States
-
Financial Services
-
100 - 200 Employee
-
Lead MLE/DS
-
Jul 2021 - Present
-
-
Senior MLE/DS
-
Apr 2020 - Jul 2021
-
-
-
Zume Inc.
-
United States
-
Packaging and Containers Manufacturing
-
1 - 100 Employee
-
Senior Data Scientist
-
Sep 2018 - Apr 2020
* Solved aggregate level and SKU level demand prediction using various models such as ARIMA, quantile regression and LSTM; solved inventory allocation problem using linear quantile regression and nonlinear quantile regression forests; reduced pizza waste up to 20% with an acceptable stockout rate * Improved safety stock model using decision tree and reduced pizza waste by up to 3% per week * Created machine learning pipeline to perform data cleaning, feature engineering, model development, and compute prediction intervals and confidence intervals for various demand prediction models * Frequently used SQL to query data and store features and predictions in PostgreSQL databases, SnowFlake, and BigQuery * Designed API in Postman and developed Flask API for various demand prediction models * Productionized demand prediction models using Flask API and Docker container; created unit test for each model; used Gitflow for version control; and deployed models via CI/CD pipeline * Created visualization dashboard in Tableau to report weekly waste and stockout metrics to managers * Worked with No-SQL data to create personalized customer profile and to create structured data for marketing activities * Implemented quantile regression for different markets in Keras using advanced features such as Functional API and custom objective function with very close results to the Statsmodels implementation * Used random forest and logistic regression to estimate likelihood of placing an order for customers * Created image annotated data and used it to identify defects using CNN Show less
-
-
-
UC Berkeley Institute of Transportation Studies
-
Berkeley, CA
-
Post-doctoral Researcher
-
May 2016 - Sep 2018
• Developed a method to estimate fuel consumption for automated truck platooning • Developed a 2D micro-simulation program in C++ for automated trucks • Developed vehicle dynamic models for automated trucks • Conducted Monte Carlo simulation of automated trucks in a large scale real world corridor • Assisted in a behavioral study of drivers who used an automated truck system on a public road • Developed a method to estimate fuel consumption for automated truck platooning • Developed a 2D micro-simulation program in C++ for automated trucks • Developed vehicle dynamic models for automated trucks • Conducted Monte Carlo simulation of automated trucks in a large scale real world corridor • Assisted in a behavioral study of drivers who used an automated truck system on a public road
-
-
-
University of Illinois at Urbana-Champaign; University of California Berkeley
-
Urbana-Champaign, Illinois Area; Berkeley, California
-
Reseach Asistant: National Work Zone Safety Grant
-
Mar 2013 - Aug 2017
• Developed a 1.5 day nationwide workshop on traffic analysis in work zones • Co-taught the course in 8 State DOTs • See the attached course webpage in National Work Zone Safety Website • Developed a 1.5 day nationwide workshop on traffic analysis in work zones • Co-taught the course in 8 State DOTs • See the attached course webpage in National Work Zone Safety Website
-
-
-
University of Illinois Urbana-Champaign
-
United States
-
Higher Education
-
700 & Above Employee
-
Research Assistant
-
Aug 2007 - May 2016
• Mastered field traffic data collection, reduction, cleaning and analysis for 5 highway work zones• Evaluated statistical effects of flagger, police, and speed photo enforcement on traffic flow• Developed statistical models (e.g. linear and nonlinear regressions) of traffic flow in work zones• Developed mathematical models to estimate capacity, queue length and delay in work zones• Proposed a method to solve “sign location” problem for speed harmonization (Large scale mixed integer nonlinear optimization)• Proposed a transformation method to optimize speed harmonization with piecewise traffic stream models• Developed a new supervised learning method to calibrate a second order hydrodynamic model Show less
-
-
Teaching Assistant
-
Jan 2015 - May 2015
I was TA for Traffic Signal Systems (CEE 517) which was a graduate course. The course introduces different types of traffic signals, queue calculation methods, coordination strategies, etc. I was responsible for grading, holding office hours and I assisted in designing midterm and final exams.
-
-
Teaching Assistant
-
Jan 2014 - May 2014
I was TA for Traffic Flow Theory (CEE515) which was a graduate course covering advance topics such as car following models, hydrodynamic models, etc. I was responsible for grading and holding office hours, and I assisted in designing midterm and final exams. I was substitute lecturer for topics such as shockwave theory, second order hydrodynamic models, and VISSIM lab.
-
-
-
Sharif University of Technology
-
Iran
-
Higher Education
-
700 & Above Employee
-
Researcher : Traffic Assignment Under Uncertain Fuzzy Travel Time (M.S. thesis)
-
Aug 2006 - Aug 2007
• Developed a method to determine the shortest path (global optimal) with fuzzy link travel times • Developed a method to solve fuzzy traffic assignment returning the global solution • Implemented the methods above in C++ for a real network with 900 nodes, 2500 links and 7157 origin-destination pairs • Developed a method to determine the shortest path (global optimal) with fuzzy link travel times • Developed a method to solve fuzzy traffic assignment returning the global solution • Implemented the methods above in C++ for a real network with 900 nodes, 2500 links and 7157 origin-destination pairs
-
-
Education
-
2016 - 2018
UC Berkeley College of Engineering
Post-doctoral researcher, Traffic and Transportation Engineering -
Univeristy of Illinois at Urbana-Champiagn
Doctor of Philosophy (PhD), Transportation and Traffic Systems Engineering -
University of Illinois at Urbana-Champaign
Master of Science (MS), Operation Research -
Sharif University of Technology
Master of Science - MS, Transportation Planning and Traffic Engineering -
2001 - 2005
Sharif University of Technology
Bachelor of Science (BS), Civil and Environmental Engineering
Community
