Jan 2023 – Nov 2023 · 10 mos · Los Angeles Metropolitan Area · Remote

• Research Assistant at Signal Analysis and Interpretation Laboratory (SAIL), USC
• Developed and tested ML pipeline to support Generative Language Models for Zero-shot Social Media Intelligence on distributed data systems using Pandas, AirFlow, PySpark & Cassandra.
• Reduced training data required for similar performance by LLMs by 90% and Improved the performance of Large Language Models like GPT3, GPT3.5, and FLAN-T5 for multi-label classification by 18%.

May 2022 – Aug 2022 · 3 mos · Seattle, Washington, United States

• From-scratch developed and tested a Python-based tool to display and analyze complex data states from ML architecture using PySpark, PyArrow & Plotly; resulting in improved performance on accuracy by 52%.
• Created and maintained ETL pipeline for tool to support multiple Large Language Models and faster visualization by 40%
• Built Python & bash scripts and worked on system pipeline performing feature extraction, data preparation, data wrangling, model selection, training, testing, and deployment with large-scale distributed data systems in high-computation environments using PySpark, Hadoop, hdfs, Docker, and Kubernetes.

Feb 2022 – Apr 2022 · 2 mos · Los Angeles Metropolitan Area

• Constructed Scalable Zero-shot Entity Linking Framework in Python on WikiData with Dense Mapping at the Centre of Knowledge Graphs for Knowledge Graph Toolkit from multiple Data sources; parallelized the data ingestion process for faster processing.
• Built and embedded APIs to Toolkit to map unseen entities in the text to WikiData nodes; improved performance of mapping by 23% using BERT-based Bi-encoders, Cross-encoders & FAISS.

Aug 2019 – Jun 2020 · 10 mos · New Delhi, Delhi, India

• Created web application for AQI Prediction and real-time Image Captioning using Angular, JavaScript, and Python; designed and tested APIs using Flask; built scripts to streamline the training and testing process.
• Engineered android Application in Java for text messaging; integrated chat-bots trained using seq2seq model, built authentication, user profile page, and custom bots based on user emotion using Firebase as a data store.

Jul 2018 – Dec 2018 · 5 mos · New Delhi Area, India

• Created Recommendation System for jobs in a team of 8; implemented Resume Parsing, Stable Matching, Collaborative Filtering, and custom NER to improve job matching accuracy from 88.12% to 94% using Python, NLTK, Scrapy & TensorFlow.
• Deployed, tested & maintained websites on GCP; created custom ontology similarity matching algorithm for improved recommendations on jobs using Protégé, OWL, and NLTK.