Building finetuned LLMs, RAG pipelines, and AI Agents for production to solve large scale problems

Researcher in Computing For Social Good (CSG)• Violence incitement detection for social media platforms advised by Prof. Mai ElSherief and Prof. David Muchlinski (EMNLP 2023)• Leading the technical development of the research project building a custom BERT model fine-tuned on a novel dataset• Prototyping alternative methods like few-shot & zero-shot learning and in-context learning using LLMs

Teaching Assistant (Spring 2023): Database Applications (CSE132B)Teaching Assistant (Winter 2023): Database Management Systems (DSE201)Teaching Assistant (Fall 2022): Beyond Relational Database (DSE250)Teaching Assistant (Spring 2022): Statistical Natural Language Processing (CSE256)

• Developed ML algorithm based mobile notification pipeline to analyze the viewers’ activity increasing engagement• Identified 80% threatening comments from 200k daily live comments by designing a robust NLP system utilizing spaCy and Gensim• Designed the algorithm for platform’s Recommendation System using collaborative filtering and extracted a novel user profile data point for it leading to 7% viewer retention increase using Fasttext, snowNLP and NLTK• Deployed models with parallel execution enhancements on GCP utilizing Docker and Kubernetes for auto-scaling capabilities• Improved the content on the app by automating the analysis of KPIs like monetization and viewer growth using Google BigQuery• Launched an HR portal that managed employees’ salary and leave records using Flask and PostgreSQL

• Identified potential customer subsets for communication using clustering and PCA leading to reduction of annual costs by $20K• Reduced churn by 17\% by creating a visualization tool for data statistics using query-based live interactive dashboards and graphs

Lead a team of 5 members to develop a currency detection system using CNNs for summer research course about Neural NetworksAcquired training in Hadoop, Mapreduce and Kafka to solve Big data problems by Hewlett Packard Enterprise

• Developed an ML based schedule-recommendation system using Surprise to optimize energy usage in home automation• Achieved 30% energy savings in a 4 person household by leveraging real time analytics of occupants’ routine

Worked on the JiLL robot as part of the Wipro tech team at JLL and also presented ideas for smart offices/workspaces.