Sep 2022 – Jul 2024 · 1 yr 10 mos · Gurugram, Haryana, India

• 1. Duplicate Image-listings Detection System
• Integrated a comprehensive duplicate listings detection system based on price, area, location and similarity score between
• Image Embeddings generated by Constrastive Language-Image Pretraining Model (clip-ViT-B-32).
• Employed multiprocessing for compute-intensive and multithreading for I/O-based tasks to optimize system efficiency.
• Achieved notable reduction in duplicate listings on the platform, improving the overall customer experience.
• 2. Fraud Detection System | SoldOut Property Detection System
• Collected listing attributes and user behavior click stream data from BigQuery using SQL queries in Databricks.
• Developed classification models XGBoost and Random Forest, achieving 86% precision and 84% recall for fraud detection.
• Implemented End-to-End pipelines for model deployment and conducted data drift experiments to maintain performance.
• Achieved 81% average reductions in fraudulent activities over the following 3 months on platform, earning recognition
• with an excellence award.

Aug 2021 – Sep 2022 · 1 yr 1 mo · Pune, Maharashtra, India · On-site

• Network simulations to predict set of potential genes of maximum influence in a biological system
• Built a multistart adam optimization model applying heat diffusion algorithm for network propagation to predict the set of most influential genes/DEG’s responsible in a disease condition using Igraph and PyTorch frameworks.
• Visualized the pathway’s nodes and their relationship in protein-protein interaction network using Neo4j DBMS with CQL query.
• Performed pathway analysis to obtain prioritized modulated signaling pathways using SPIA algorithm.
• Extraction of key entities from pdf documents related to scientific domain
• Built hybrid system of Biomedical NER models and heuristics for extracting various variables of interest to the client.
• Developed NLP based pipeline to extract entities from pdf drug labels using Poppler, Stanza, Spacy, Elastic Search, and RegEx.

May 2020 – Jul 2020 · 2 mos · Kharagpur, West Bengal, India

• Multimodal Tweets Analysis for disaster response using Computer Vision and NLP
• Built multimodal deep learning architectures using pre-trained computer vision and NLP models such as Resnet50+AWD-LSTM, Resnet152+XLNet, Vgg19+XLM, Densenet161+BERT, and various other combinations.
• Implemented Intermediate, Late, and EmbraceNet fusion techniques to combine the image and text modalities.
• Applied discriminative fine-tuning, slanted triangular learning rate, and gradual unfreezing techniques.
• Obtained F1-score 88% and 85% on binary and multiclass classification tasks with the improvement of 3% and 4% respectively over the performance of models with single modalities.

May 2019 – Jun 2019 · 1 mo · Kharagpur, West Bengal

• Prediction of occupational accidents according to activity type using ensemble methods and optimization techniques
• Applied Machine Learning models - Logistic Regression, Decision Tree Classifier, K-Nearest Neighbors, Support Vector Classifier, Random Forest Classifier, Gradient Boosting Classifier and ensemble techniques like Bagging, AdaBoost and Simple Voting Classifier.
• Implemented single-objective optimization techniques namely Simulated Annealing and Differential Evolution using the accuracy of ensemble model as the objective function and aslo applied multi-objective optimization techniques like Non-dominated Sorting Genetic Algorithm (NSGA) and Multi- Objective Differential Evolution (MODE) using total precision and total recall as the objective functions.
• Generated new data using the trained ensemble model which was then used to extract safety rules to prevent further accidents.

Aug 2018 – Apr 2019 · 8 mos · IIT Kharagpur, India

• Coordinated sponsorship drive in Mumbai and Kolkata raising INR 8.8 lakhs by Alumni Contribution.
• Oversaw the accommodation of 4500 participants by leading a team of 8 Associate members along with crew members.
• Co-headed a team of 35 Associate members, expanding outreach of Spring Fest to 450+ colleges in India and Nepal.
• Conducted Eliminations of Poetry Slam and Hilarity Ensues in Brewberrys Bhubaneswar, which had a footfall of 400+people.
• Supervised the publicity campaign in Uttar Pradesh and Uttarakhand, published 18 media articles in different regional and
• national newspapers and magazines in India, 3 times YoY increase, also increased participation by 26.1%.

Aug 2018 – Dec 2018 · 4 mos · IIT Kharagpur