May 2019 – Jan 2021 · 1 yr 8 mos · Gurgaon, India

• Leading a team of junior engineers and mentoring them on ML & software engineering best practices.
• Project and stakeholder management.
• Churn modeling to predict when a cardholder will become inactive. Achieved AUC of 0.8 & capture rate of 70%. System design and data pipeline to process TBs of data, scoring billion plus cards.
• Visualizing Covid spread using transaction data. This gained traction from CXOs, resulting in new product areas, and collaborations with Governments \& research institutes.
• Automated Chargeback Dispute Resolution process using OCR, NLP, and rule engine. Reduced process time by 90%.
• Envisioned and built dashboards to deliver insights into chargebacks, disputes, \& frauds. Commercialized as a product for banks.
• Filed patents in card inactivity modeling, and tracking disease spread using transaction data.

Sep 2017 – Dec 2018 · 1 yr 3 mos · Noida Area, India

• Joined back Info Edge after 2 years away doing masters. This time as an ML Engineer/Technical Architect.
• Solving business problems for leading Indian portals in job search (naukri.com), real estate (99acres.com),
• and education (shiksha.com).
• Functioning as the lead engineer and product manager.
• Developing new data-driven products leveraging the rich data available.
• Provide insights to improve products and data quality.
• Projects:
• House Price Prediction: Built a meta learner build on top of different ML models. Responsible for the entire life cycle of the project including product idea, data analysis, modeling, API creation and deployment via docker.
• Similar Entity Search: Developed a general-purpose entity search system based on representation -
• learning and similar neighbor search. Entities could be images, documents, or any objects having vector representations.
• Duplicate Image Detection: System to flag accounts using celebrity pictures & multiple accounts using the same profile pictures. Used siamese neural networks
• Similar Question Search: Used doc2vec and nearest neighbor search to find similar questions
• Salary Prediction

Jan 2017 – Jun 2017 · 5 mos · Stockholm, Sweden

• I am doing my master thesis with the Machine Intelligence group at Ericsson Research.
• The thesis is on "Anomaly Detection for Temporal Data using Long Short-Term Memory (LSTM)".
• I am developing algorithms for anomaly detection and predictive maintenance using LSTMs. The work also involves automating machine learning workflows using Bayesian optimization.
• Thesis is being done as part of the 5GEM (5G Enabled Manufacturing) project involving Ericsson, SKF, and Chalmers University ofTechnology, Gothenburg.
• . Thesis Link: https://www.diva-portal.org/smash/record.jsf?pid=diva2%3A1149130&dswid=3525

Aug 2016 – Aug 2017 · 1 yr · Stockholm, Stockholm County, Sweden

• Pursuing second year of masters at KTH, Stockholm. Specializing in Data Science for Big Data. Doing master thesis at Ericsson Research, Stockholm.

Aug 2015 – Jul 2016 · 11 mos · Eindhoven, North Brabant, Netherlands

• First year of master program at TU Eindhoven, Netherlands. Major in Data Science and Algorithms, minor in Technology Entrepreneurship.

May 2013 – Jul 2015 · 2 yrs 2 mos · Noida

• A part of 99acres search and analytics team. Tasks include taking care of 99acres search system, data analysis, modeling, machine learning, and data visualization. Technologies that I work on include JAVA, R, Apache Solr, MongoDB, MapReduce. Some of the projects I worked on include:
• Search Ranking Framework: This involved building a Java based framework for computing search rank scores of real estate listings for 99acres.com. The search rank algorithm balances buyers' and sellers' concerns with the monetization needs of the business. This involved modeling click-response and listing quality, formulating lead demand/generation, defining metrics, A/B testing etc. The project improved search revenues by 20%.
• Click-Through Rate Prediction: This project involved deriving a power-law equation for predicting click-through rate of a listing. This was done by applying linear regression to log-transformed data. R was used for modeling and MongoDb map-reduce was used for computation.
• 99acres Search: Implemented 99acres' search using Apache Solr, Named Entity Recognition(NER) and Spell-Checking to convert a free-text query to a parameterized search query.
• Home Image Classification using CNNs
• Spam Filter: Built a model to highlight malicious listings, thereby reducing such postings by 60%.

Mar 2012 – May 2013 · 1 yr 2 mos · Hyderabad Area, India

• Developer in Oracle Fusion Product Information Management (PIM) Team.
• Domain : Supply Chain Management.
• Developed new features for the product Change Management module.
• Worked closely with product manager and QA as part of agile team
• Technologies : Java , ADF , Linux , SQL , WebLogic.

Sep 2010 – Feb 2012 · 1 yr 5 mos · Noida

• Worked as full stack java developer in projects for clients American Family Insurance and New York Life Insurance.
• Develop and Owned Software modules.
• Built POCs for technical evaluation.
• Unit Testing.
• Technologies : Java ,Struts , XSLT , SQL , Java Script, CSS, HTML.

Jan 2010 – Jun 2010 · 5 mos · Noida Area, India

• Trainee In Soft Remote Development Team.
• Soft Remote is an implementation of the IPSec Protocol.
• Technologies : C++ , IPv4/IPv6Socket Programming, Monitoring utilities like wireshark, tcpview etc.