Apr 2023 – Apr 2024 · 1 yr · Bengaluru, Karnataka, India · Hybrid

• Created clusters of past doubts for 200 chapters of IIT-JEE and NEET using OpenAI model embedding and Quality Threshold Clustering Algorithm with hyperparameter optimization.
• Implemented a functionality to recommend maximum 4 similar past doubts based on semantic and contextual meaning which could decrease the funnel by 45%.
• Integrated a contextual doubt classification model built using semi-supervised learning technique and Logistic Regression, which improved product metrics by 40%.
• Delivered an API endpoint using FastAPI to recommend similar doubt in real-time, fast, and scalable way, reducing daily cost and Turnaround time from 24 hours to instantaneously.
• Executed A/B testing methodologies leveraging statistical technique, while upgrading recommendation model

Jul 2022 – Mar 2023 · 8 mos · Bengaluru, Karnataka, India · Hybrid

• Developed fast, efficient recommendation model for short educational videos using Bayesian Ranking Algorithm, which solves Explore-Exploit dilemma. deployed pipeline on Airflow resulted in 0.71 NDCG score.
• Delivered API endpoint to return personalized ranking in real-time, fast and scalable for serving for 2M users.
• Created a doubt classification model into academic and non-academic categories with an accuracy of 92% using text processing, TF-IDF, Random Forest Classifier and delivered an API endpoint to integrate into product.
• Explored Reinforcement learning to do POC on cross-selling products.

Jan 2022 – Jul 2022 · 6 mos · Gurugram, Haryana, India

• Mentored a team of 4 members and part of an Interview panel in the Technical Hiring team.
• Image Classification of a patch of human scalp & predict the count of hairs in 3 predefined ranges.
• Analyzed and preprocessed 600 images of 3 classes using OpenCV. Build Convolutional Neural Network Model (CNN) with transfer learning and got an accuracy of 76% in Keras, TensorFlow.
• Increased accuracy with pre-trained ResNet34 CNN architecture to 78% and also implemented Vision Transformer and MLP Mixer architecture to improve classification metrics by PyTorch.
• Created 5 production jobs in Kubeflow environment, collaborated with business team to continuously enhancing rule-based recommender system based on the requirements.

Nov 2020 – Dec 2021 · 1 yr 1 mo · Gurugram, Haryana, India

• Developed Machine learning & ANN models to minimize churn rate as an internal project of the company in Telecom, BFSI domain and performed EDA, feature engineering, feature selection, modeling, hyperparameter tuning, and feature importance to reduce features.
• Generated new datasets by combining data from multiple source systems with SQL queries and completed feature engineering on data having 100+ columns.
• Developed a mathematical optimization model using Gurobi in Python to find decision variables that result in an increase in Net Revenue by 8% & delivered one monitoring dashboard to track business KPIs in Tableau.

Jul 2018 – Apr 2019 · 9 mos · Bengaluru Area, India · On-site

• Designed and integrated self API onboarding feature to add/update backend link of APIs with SQL queries to streamline the process and reduce two days of team effort monthly in Java and Python.
• Completed unit testing & maintained code changes in Git Repository.