Apr 2024 – Nov 2024 · 7 mos · Bengaluru, Karnataka, India

• Improved SmartCopy: A copy optimization model to predict the copy variant that members are most likely to engage with.

Dec 2021 – Apr 2024 · 2 yrs 4 mos · Bengaluru, Karnataka, India

• Lead the LinkedIn Groups Creator AI charter, collaborating with product and engineering teams.
• Designed and implemented a content-to-group recommendation system that achieved a 67% site-wide impact on Groups Creators with Response, currently under patent review (Patent Pending: Group Recommendation based on Topic Clustering).
• Built a question-answering framework leveraging generative AI for LinkedIn Groups (Patent Pending: Question Generation for Groups and Communities).

Feb 2020 – Nov 2021 · 1 yr 9 mos

• Developed a data extraction framework for unstructured documents. (Patent Granted: WO2022040464A1)
• Developed multiple named entity recognition models to anonymize data. These models eliminated major challenges to enterprise data sharing.
• Lead the effort in building a data version control pipeline using MLFlow.
• Lead the architecture of RabbitMQ and ZeroMQ based message queue systems to replace all REST-based ML services.
• Revamped the MLOps pipeline by moving all machine learning microservices to docker containers.

Mar 2019 – Feb 2020 · 11 mos

• Built a data extraction framework that can extract key-value pairs from text without any data-specific fine-tuning (Patent Pending: US20210064821A1).
• Researched and deployed multiple text classification models for email triage.
• Created an internal text classification library to facilitate adding new classifiers to production.
• Built a question answering system that can answer free text queries by performing a smart document search.

Aug 2017 – Mar 2019 · 1 yr 7 mos · Pune, Maharashtra, India

• Built a text classification system using transfer learning to classify customer feedback into various categories.
• Built look-alike models to identify similar customers based on their purchasing patterns and demographic information.
• Created a data-driven buy-plan for the apparel business unit.

Apr 2016 – Aug 2016 · 4 mos

• Worked as an open-source developer under the Python Software Foundation.
• Added Gaussian distribution for Bayesian Networks in Pgmpy, a python library for machine learning using probabilistic graphical models.
• Code reviews and Unit tests were required for each commit. Gained hands-on experience with python, and libraries like numpy and pandas.