➜ Contributed in development of image-based models which (a) identifies size chart containing images and, (b) extracts size charts from product images for products where size charts are relevant. This model is live for 13 cross category product types in US marketplace and is estimated to have an impact of $1.1B annualized OPS and $475M annualized net concession opportunity➜ Productionized NLP models to extract attribute value from unstructured text such as title, bullet points and product… Show more ➜ Contributed in development of image-based models which (a) identifies size chart containing images and, (b) extracts size charts from product images for products where size charts are relevant. This model is live for 13 cross category product types in US marketplace and is estimated to have an impact of $1.1B annualized OPS and $475M annualized net concession opportunity➜ Productionized NLP models to extract attribute value from unstructured text such as title, bullet points and product description at scale (135 product types from 8 marketplaces). This program is estimated to make an impact of $640 M for these 135 product types Show less

➜ Developed customized strategy to assess effectiveness of existing promotions for an American-German hair care giant. Implemented a regression model to distinguish between baseline and promotion related incremental quantities. This allowed us to asses an uplift of 39% and helped compute ROI at 22%➜ Developed Application Scorecards (A-Score) for the small and medium-sized company (SMB) segments for one of Indonesia's top 5 banks Performed cohort analysis of vintage curve & roll rate graphs… Show more ➜ Developed customized strategy to assess effectiveness of existing promotions for an American-German hair care giant. Implemented a regression model to distinguish between baseline and promotion related incremental quantities. This allowed us to asses an uplift of 39% and helped compute ROI at 22%➜ Developed Application Scorecards (A-Score) for the small and medium-sized company (SMB) segments for one of Indonesia's top 5 banks Performed cohort analysis of vintage curve & roll rate graphs to define a bad loan and implemented logistic regression model to predict loan approval using demographics, banking, bureau and financial data with a precision of 82% and recall of 63% Show less

Working in Network Operations and Customer Analytics Team➜ Engineered a new customer focused KPI, Customer IRROP Index (CII) which measures how good or bad a day was for our customers based on 11 operational factors.➜ Productionized a machine learning powered solution to benchmark Customer IRROP Index (CII) which gives insights around operational performance improvement.➜ Implemented CatBoost Regressor (python) to predict Taxi-In Time which when added to Estimated-On Times gives… Show more Working in Network Operations and Customer Analytics Team➜ Engineered a new customer focused KPI, Customer IRROP Index (CII) which measures how good or bad a day was for our customers based on 11 operational factors.➜ Productionized a machine learning powered solution to benchmark Customer IRROP Index (CII) which gives insights around operational performance improvement.➜ Implemented CatBoost Regressor (python) to predict Taxi-In Time which when added to Estimated-On Times gives Estimated Arrival Time, Replaced Flight Aware, saving of $500K annually➜ Validated linear relationship between operational KPI D:00 (departure on time) with Boarding Time and leveraged this relationship to improve D:00 by modifying Boarding Time➜ Led Fuel Efficiency effort (pyspark), developed database of optimized routes by finding potential short-cuts between origin and destination, extensively used by Network Planning for scheduling purposes➜ Engineered automated pipelines (python/ms-excel) for running ETL and Reporting jobs➜ Led the use of GitHub for source-code management and active collaboration➜ Using NLP (python: gensim), estimated brand sentiment value from customer satisfaction surveys Show less

Working in Fractal Products division: Eugenie➜ Led forecasting requests from MARS Incorporated and Procter & Gamble at Proof of Concept (PoC) stage using data driven models (python)➜ Engineered real-time anomaly detection solution that proactively prevents server down times (pyspark streaming)➜ Revamped algorithms written in pandas to its corresponding pyspark version for improving scalability

➜ Engineered Hadoop Cluster of 3 virtual machines on CentOS➜ Implemented python code to refactor third-party data into Hadoop cluster along with encryption/decryption➜ Created synthetic data of GST forms and simulated synthetic data with a python script (HappyBase) using Thrift API for read/write with HBase

➜ Detailed study of the exploitation of vulnerabilities, incident investigation, and preventive techniques.➜ Setup of a virtual environment for simulation of attacks (Kali Linux, a website using Drupal, Windows XP/7).➜ Simulated two exploits: CVE-2014-3704 and, CVE-2010-1240 (using Kali Linux). These two exploits are used for web defacement and phishing attacks. ➜ Simulated WannaCry Ransomware attack using ETERNALROMANCE exploit (payload was executed using Kali Linux Metasploit… Show more ➜ Detailed study of the exploitation of vulnerabilities, incident investigation, and preventive techniques.➜ Setup of a virtual environment for simulation of attacks (Kali Linux, a website using Drupal, Windows XP/7).➜ Simulated two exploits: CVE-2014-3704 and, CVE-2010-1240 (using Kali Linux). These two exploits are used for web defacement and phishing attacks. ➜ Simulated WannaCry Ransomware attack using ETERNALROMANCE exploit (payload was executed using Kali Linux Metasploit framework). Successfully, compromised and encrypted the target machine (Windows XP/7). Show less

➜ Worked with E-filing project team, the team consisted of 5 members. I was responsible for form validation schema.➜ Form validation using MS-Excel.