Sep 2022 – Dec 2022 · 3 mos · Gurugram, Haryana, India · Remote

• Designed machine learning models to estimate the probability of train tickets getting confirmed.
• The models reached a precision of 96% and a recall of 92%.
• The non-confirmed tickets are guaranteed a flight ticket against a small premium. The models created an overall positive expected return for the company.

Aug 2021 – Dec 2022 · 1 yr 4 mos

• Accounting software
• Designed the system to manage accounting and inventory for the small and medium scale businesses.
• The product on-boarded 10 businesses having 750-1000 clients each.
• Aptology
• Designed the system to screen candidates for a role by evaluating the survey filled by them. It analyses their personality traits and finds the best fit.
• It has been used for more than 100 job profiles and 10000 candidates.

Sep 2019 – Jul 2021 · 1 yr 10 mos

• Got admitted to the PhD program with the top research funding organization in France (CNRS) on a full scholarship.
• Performed two years of studies before exiting the program due to COVID-19 impact.
• The focus of my thesis was to develop dynamic networks where the number of layers / architecture in a deep learning network can vary dynamically according to the difficulty of the data (images and videos) to be processed.
• Devised a gating mechanism that could select different parts of the convolutional neural network (CNN).
• Experimented extensively and achieved a relative improvement of 4% on the classification task.
• Achieved a relative improvement of 10%-12% on depth inference task from a single image.

Aug 2017 – Aug 2019 · 2 yrs

• State-of-the-Art-Report
• Published a report to summarise around 500+ research papers for a variety of applications such as object classification, detection and improvising techniques such as self-supervised learning, online learning, etc.
• The report has been cited 90+ times.
• Original Research Work
• Enhanced the existing state-of-the-art deep learning architecture on object detection to improve performance on datasets with fewer training data by 3% in absolute terms.
• Came up with the idea, experimented a wide variety of configurations and wrote the resulting paper.
• The paper was accepted in ICPR 2020.

Jul 2014 – Jul 2016 · 2 yrs · Bengaluru, Karnataka, India

• Addition of new features:
• Built new back-end libraries in C, C++ for the rail product.
• Added new features in the rail product - Add Complex Itineraries, Cancel Complex Itineraries, Cancel Per Pax, Open Ticket, Split Per Party, Seat And Berth Preferences, Immediate Distribution of Tickets, VAT Exclusion, Resend Tickets and New Supplements.
• Addition of new client:
• Part of the design team to add a new client for Rails product.
• Added enhancements for method of deliveries, corporate codes, seat merchandising, fares, special supplements, cancellation policy and purchase conditions.
• Later worked on adapting the APIs with the new client's infrastructure.
• API migration:
• Migrated Rails back-end for Shopping, Booking, Ticketing and After Sales to the new client's infrastructure.
• Updated all the APIs as well as conducted black-box testing in all pre-production and production environments.