Experience

Nilos

• Israel
• Financial Services
• 1 - 100 Employee

• Software Engineer

  • Mar 2022 - Present



ShopperAI

• Israel
• Software Development
• 1 - 100 Employee

• Computer Vision Engineer

  • Jun 2021 - Apr 2022



Via

• United States
• Software Development
• 700 & Above Employee

• Data Scientist

  • Oct 2019 - Jul 2020

  ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬ Rider Reports Classification: - Business Problem refinement: refining the label categorization according to business problematics. Categorizing across: vehicle issues, pickup-dropoff issues, driver complaints, safety issues. - Data Acquisition: Acquiring data from SQL database, and integrating into Pandas DataFrame. - Preprocessing: joining between multiple tables, data cleaning: missing values, incorrect labels. - Data Exploration: exploring whether dataset is balanced / unbalanced, correlation between labels, matplotlib, seaborn. - Data Labellization of training data: Jupyter Notebook, ipyWidget, labellization by students of about 3k rows. - Model Training: splitting data across train/test subsets. Researching most relevant model towards training. Choosing SotA Transformers DistilBERT NLP model to extract word embeddings of each words and sentence in the Training data. Loss function: sum of logits across category probabilities. Saving to a PyTorch model with Pickle. - Model evaluation: accuracy metric. Analyzing confusion matrix to pinpoint model issues, and iterate again with newly labelled more specialized data. - Deployment: creating a new column in the SQL database with the more precise category detected by our model. Working together with Devops Engineer in the team to load dynamically the categorization of issues in a live map based on Google Maps API ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬ Building a live map centralizing operations issues, such as off-route, pickup/dropoff delays, client complaints. Normalizing delays, and attributing score to road portions to indicate a "delay indicator". Correlating those with rider complaints to identify most problematic portions. Defining custom metrics such as off-route count, sum of delays, delay cost Serving with REST API based on AWS Lamba. - Inputs: city, which metrics, count_limit - Business logic: querying SQL database securely - Output: JSON with categories, metrics etc. Show less



Israel Tech Challenge <itc>

• Israel
• Software Development
• 1 - 100 Employee

• Student Mentor

  • Oct 2018 - Sep 2019

  Member of the training team with 6-7 other peers. Reporting to Program Director. Responsibilities: - Mentoring students - Making sure students are on track with academic workload and identifying academic difficulties or exceptional performance - Mentoring Data Science projects: about 50 hours project giving students exposure to real-life end-to-end Data Science project encompassing the whole Data Science lifecycle (Business Problem definition, Data Acquisition, Preprocessing, Model Training, Model evaluation, Deployment) - Correcting and evaluating exercises in Python, Statistics, Probability, Data Modelling - classical ML: i.e - Random Forest, K-Means, SVM, Logistic Regression) Show less



Check Point Software Technologies Ltd

• United States
• Computer and Network Security
• 700 & Above Employee

• Data Scientist

  • Apr 2018 - Sep 2018

  Clustering of malwares' families on executable files Improvement of malware classification on pdf files Clustering of malwares' families on executable files Improvement of malware classification on pdf files