May 2022 – Aug 2023 · 1 yr 3 mos · Bengaluru, Karnataka, India · On-site

• 1. Booking-Fraud-Prevention-Realtime: Revamped a 2.5-year-old regional model’s performance, with SPAUC@0.01 jump from 0.82 to 0.94. Trained the model on the latest fraud MO.
• With more than 6 countries the model is live in, it has reduced the losses by more than half in 3 countries.
• For other countries, minimized the gross loss to satisfy the business tolerance, friction & decline rate.
• 2. Fraud-Tagging: Launched the base model to help the Risk-Ops team to tag the chargeback cases as Fraud or Friendly, to help dispute management.
• With a SPAUC of 0.01, reduced the manual work hours by 90.
• 3. Topup-Fraud-Prevention-Realtime-v6: Revamped the model to capture Topup frauds in realtime.
• Previous versions were trained on old and sparse data.
• Setup an automated end-to-end pipeline with data preparation, training, inferencing, and serving.
• Introduced additional labels to the training pipeline, to capture a wider profile of frauds.
• Increased the AUC-ROC from 87% to 96% and SPAUC from 71% to 86%.
• 4. Enhanced-Feature-Bank-v2: Rebooted and reconstructed a data repository that serves the base data for multiple models in Production.
• Owing to changes in the ecosystem, many critical features in the data bank had atleast 99% of the values null.
• Filled the gap by co-joining with new data, and reduced the gap by atleast 33% to 60% values.
• 5. Fraud Cases Investigation:
• 1. Booking Fraud Spike: Discovered a bug on the backend that induced frequent fraud spikes & an increasing decline rate for past 4 months.
• 2. Model Rollout Issue: Identified the root cause behind a major fraud spike. A highly efficient model (able to prevent around 90% frauds) was launched for only 10% traffic over a span of atleast 2 years.
• 3. Risk Ops Explainability: Automated Pipeline aiming to supplement the dashboard with crucial data that enables the ops team to better investigate fraud cases and suspicious users.

Sep 2020 – Apr 2022 · 1 yr 7 mos · Bengaluru, Karnataka, India

• 1. PINTEREST++:
• ∗Furniture Recognition and Detection: Developed using YoloV3. It detects and recognizes the furniture products in the customer shared images from different platforms like Pinterest, Google Images etc.
• ∗Furniture Recommendation: Furniture from the internal catalog with similar properties to that of recognized furniture is recommended to customers (using custom Deep Convolutional Neural Networks, Computer Vision).
• ∗Increased Spacejoy’s e-commerce conversion ratio and non-design e-com orders.
• ∗Increased the adaptation of customer preferences in the room designs by the Spacejoy designers.
• ∗Deployed APIs to serve multiple functionalities (like showing recommendations for products in the box drawn by customers) and set up the segregated infrastructure to reduce response time.
• 2. Developed Product Recommendation Engine, for the internal designer team and Spacejoy customers to improve the shopping experience and e-commerce conversion.
• 3. Automated the whole Analytics and Reporting, to enable a consistent information/data flow throughout the organisation.

Jan 2019 – Jun 2019 · 5 mos · Gurugram, Haryana, India

• 1) Designed an algorithm for "Detection of Duplicate Products in an E-commerce Catalog". With more than 6 million products on the catalog, the algorithm was able to detect all the duplicates of a product on an average in less than 1 second.
• 2) Developed a chatbot Ms Noona for the HR department.
• 3) Precisely identified the root cause in the Search Engine Implementation on Algolia causing the poor search results. Lead to the abolishing of Algolia for further use, benefitting the company in monetary as well as in operational terms.

Mar 2018 – Sep 2018 · 6 mos · Patiala, Punjab, India

• 1. Headed the college club for a duration of 7 months.
• 2. Lead a team of around 70 students organizing various events, workshops for students of the university.
• 3. Set up long term goals for the club, with the help of all the members also set up a roadmap for the same.

Aug 2017 – Feb 2019 · 1 yr 6 mos · Patiala, Punjab, India

• Co-founded a new university club Global Internship Council.
• 1. The main objective of the society was to help students secure international internships.
• 2. Built a team of around 50 members from the ground up.
• 3. Looked after the initial functioning of the club setting up different functional teams, long term goals and looking after the initial operations of the club.

Aug 2017 – Feb 2018 · 6 mos · Patiala, Punjab, India

• 1. Part of the technical team of the university's largest technical club for Computer Science students.
• 2. Helped organize technical events for the university's students.