Work Experience

- “Large-scale Recommender Systems and Ranking at JustEat Takeaway.com”
  Senior Data Scientist (2019 - Present)
  - Working on building prototypes for personalized restaurant and dish recommendations in order to rank items according to customer’s preference
    - Built and productionalised implicit matrix factorisation for personalised restaurant recommendations using SageMaker.
      - Used to replace popularity based baseline in personalized restaurant emails
    - Built a POC for personalised restaurant recommendations using Factorisation Machines
      - Attempt to replace strong personalized restaurant recommendations baseline on multiple channels
    - Participated and presented in recommendation meetups inside the company, which dealt with problems related to various biases and unbiased learning to rank.
    - Literature review on counter factual offline evaluation using multi-armed bandits and inverse-propensity metric.
  - Building a relevance search prototype using Learning to Rank (LTR) principles
    - Query Expansion using item2vec models
    - Rank boosting using position-bias corrected click-through-rate
    - Well versed in learning to rank loss functions
    - Involved in A/B testing newly introduced features
    - Tools Used: Elastic Search
- “Building large scale recommender sytems leveraging implicit feedback”
  Ph.D. student at Laboratoire d’’Informatique de Grenoble (LIG), under Prof. Massih-Reza Amini and Assoc Prof. Charlotte Laclau (2015 - 2018)
  - Worked under the supervision of Prof. Massih-Reza Amini and Post. Doc. Charlotte Laclau in order to apply deep learning methods to improve top-k recommendations on implicit feedback
    - Tools used: Tensorflow
    - More details here: https://arxiv.org/abs/1705.00105
  - Contributed a dataset consisting of click logs of users of Kelkoo for recommender system community
  - Set benchmark results for major recommender system baselines meant for implicit feedback on the above mentioned dataset
    - More details on: https://dl.acm.org/citation.cfm?id=3080713
  - Working in collaboration with engineers at Kelkoo and Purch on the project Calypso where the objective is to predict the probability that user clicks on a given offer using Field-Aware Factorization Machines (FFM) and feature engineering.
    - Data pre-processing in SPARK
    - FFM implementation taken from https://www.csie.ntu.edu.tw/~cjlin/libffm/
  - Attended RecSys 2016 held in Boston and RecSys Summer School 2017 held in Bolzano
  - Got papers accepted at A* conferences (SIGIR) and core-A journal (TKDE). More details here: dblp, Google Scholar
- “Mining health and nutrition information from Twitter over time”
  Research Intern at Laboratoire d’’Informatique de Grenoble (LIG), under Dr. Sihem Amer-Yahia (May 2015 - October 2015)
  - Worked in collaboration with Profs. Marianne Clausel and Massih Reza Amini to extend LDA with hiddden temporal variables in order to discover seasonal diseases in Twitter. More details here: https://ieeexplore.ieee.org/document/8263414/
  - Ported the data pipeline developed in the first internship onto ShareInsights, a robust data manipulation infrastructure
  - Found general-purpose concepts like retail products and diseases inside French tweets and statistically relating them
- “Mining health and nutrition information from Twitter”
  Research Intern at Laboratoire d’’Informatique de Grenoble (LIG), under Dr. Sihem Amer-Yahia (September 2014 - December 2014)
  - Worked in collaboration with computer scientists and geographers in the context of the CNRS MASTODONS CrowdHealth project where I developed:
    - A database indexing module in Postgres to optimize tweets extraction in real time
    - A tweet annotation module based on crowdsourcing:
      https://crowd4u.org/en/projects#p5
      http://nutritionunleashedforus.com/sidana/user/nutritionCrowd.html
    - An SVM classifier module based on a 10-fold cross validation
    - A Gibbs sampler module for inferring health information via sophisticated bayesian modelling
    - Developed a geo-based visualization interface
  - Project details available at http://slide-apps.imag.fr/crowdhealth/
- “Feature engineering and implementation of machine learning algorithms at Xurmo Technolgies Bangalore ”
  Research Engineer at machine learning based product-based start-up Xurmo Technologies,Bangalore (June 2013- November 2013)
  - Studied and wrote jobs for filtered and wrapper based feature selection methods
  - Algorithms implemented - Information gain, Chi-Square, Principal Component analysis and forward feature selection algorithm (in java)
  - Ran the algorithms on many UCI-repository datasets.
  - Results were at par with popular data mining tool - Weka.
- “System Engineer at Estel Technologies ”
  System Engineer at Global Support Team at ESTEL TECHNOLOGIES PVT LTD Gurgaon, India (February 2009 - March 2010)
  - Configuration, monitoring and trouble shooting of Mail Server
  - Deletion, assigning Quotas and and providing support for Nokia Siemens Networks users
  - Configuration monitoring and troubleshooting of FTP
  - Manage clients helpdesk for telephonic and mail Support
- “Website and application development”
  Application developer at E-Ware TECHNOLOGIES PVT LTD Gurgaon, India (October 2008 - December 2008)
  - Worked on ASP.NET infrastructure and learned in’s and out’s of visual studio
  - Obtained familiarity with common language run-time and just-in-time compilation
  - Worked on MySQL on backend

Sumit Sidana