- 2021.01 - Present
University of Maryland, Baltimore County
Skills: PyTorch, Python, Hugging Face, Scikit‐Learn, LLM, SQL, Spacy, NLTK, Pandas, Numpy, Matplotlib, Seaborn, Weight & Biases, Hydra
- Spearheaded a novel, hierarchical, semi-supervised event modeling framework, achieving up to 8.5% improvement over prior state-of-the-art approaches in 2 datasets and across 4 evaluation metrics. (Published on *SEM 23, ACL)
- Pioneering first-of-its-kind multimodal counterfactual generation dataset (8k+ real-life events), merging text and images for nuanced alternate timeline, a novel contribution to counterfactual reasoning and multimodal real-life event understanding.
- Collaboratively developing a Graph Convolutional Network for event understanding and reasoning on 2 complex datasets (250k+ data-driven event graphs), advancing research in the field of graph-based machine learning.
- Mentoring and providing research guidance to an undergraduate who is a member of an underrepresented group in CS.
- 2019.04 - 2021.01
- Multi-factor Authentication
- Backend posts caching using Redis
- Implemented Google ReCaptcha to prevent bots and frauds
- Worked with elixir to implement a scalable chat messenger
- Integrated kickofflabs for referral contest
- Integrated sendgrid for 3rd party mailing service
- Integrated ActiveCampaign for campaign management and marketing
- Integrated OneSignal for push notification in iOS, Android & Web
- Fixed a good number of Backend and Frontend issues
- 2019.01 - 2021.01
Tech Stack: Python, Node.js, React, PostgreSQL, GraphQL, AWS Lambda, AWS Lightsail
- Single-handedly grew e-commerce to 1000+ active users and $40000/month orders.
- Achieved a successful startup exit, securing a $50,000 sale, demonstrating strategic acumen.
- 2018.10 - 2019.03
Skills: PyTorch, Keras, Hugging Face, Large Language Model, NLTK, Python, MySQL, Elasticsearch, AWS EC2, Node.js, React
- Boosted the sale by ≈23% by developing a product recommendation system using Product2Vec embedding.
- Engineered a Chatbot combining AI algorithms with logic-based if-else, decreasing response time by ≈1 hour.
- Decreased server cost by ≈10% by implementing an AWS Lambda-based ML pipeline for online learning.
- Build a model to predict the dimension of a product from the known datasets which helped the traveller team to allocate the luggage by ≈20% more efficiently.
- Implemented a microservice to refresh the inventory every 12 hours which is later fetched by Facebook Ads, decreasing marketing labour by ≈10%.
- Implemented super menu on the e-commerce site to categorise products which helped ~60% users to find the right product more quickly.
- Give a ≈110% better user experience on search using word embeddings.
- Handled whole Facebook ads system, regarding this implemented many scripts and microservices
- Implemented curation scripts to curate data from Amazon, Ebay or other US based e-commerce sites.