Course: Natural Language Processing for Social Sciences

Course: Applied Deep Learning

- Tutored undergraduate athletes at Columbia College in "Calculus-based Introduction to Statistics."- Adapted teaching methods to fit diverse learning styles, ensuring a strong grasp of the material and supporting their academic progress.

Course: Analytics in Practice- Supported six teams of 4-5 students working on data-driven projects (NLP, visual dashboards, computer vision) in retail, insurance, and finance sectors.- Maintained communication between project sponsors and teams, ensuring smooth project progress.- Facilitated understanding of project expectations between the professor and teams.- Assisted teams in resolving technical and logistical issues to help address industry challenges.

Course: Advanced Analytics Techniques- Held regular office hours to support students and address queries.- Led labs/recitations to guide students in applying course techniques like advanced regression and NLP to real-world data.- Graded assignments and provided constructive feedback to ensure student understanding.

Engineered a Retrieval Augmented Generation (RAG) pipeline on AWS Sagemaker using AzureGPT API and LangChain to help executives prepare for earnings calls by querying extensive quarters of financial reports.Optimized the RAG pipeline with advanced self-query metadata filtering and integrated hybrid search functionalities in OpenSearch, significantly enhancing information retrieval accuracy.Built a Multi-Agent RAG framework from scratch using LangGraph and function calling, expanding knowledge management capabilities to support product comparisons and multi-faceted business queries.

Led a 20-member team in crafting pricing strategies for convenience stores across 17+ EU and US markets, while enabling skill growth and maintaining timely deliveries.Utilized k-NN and Random Forest for store clustering, while utilizing unobserved component modeling and constrained regression to estimate baseline sales and price elasticities, informing price recommendations.Developed and implemented code-based strategies to address open-ended problems, ensuring expedited and robust deliveries with up to 80% reduction in process turnaround time.Built automation pipelines on Azure and AWS for efficient vendor data collection and insight generation, resulting in complete process optimization and facilitating Accenture's business expansion with the client.Awarded Accenture Star Award for demonstrating technical expertise in creating data-driven solutions.Achieved 3rd rank in Accenture Global Hackathon among 80+ teams across 15+ countries.Certified Accenture's Top 2% Data Scientists globally through Workera assessments.

Structured an open-ended sales problem and developed a robust predictive model for near real-time tracking of the conversion probability of a sales opportunity. Formulated menu-level price recommendations analyzing survey data for a major food chain spread over 5+ markets. Curated and automated 4 stage-wise reports to ensure error-free and structured analysis of recommendations. Built a production pipeline for the same. Reduced turnaround time of the overall pipeline by 20%.

Built robust data-driven models for real-time and accurate estimation of State of Charge (SoC) and State of Health (SoH) of Li-ion Batteries subjected to 1. Constant Current Constant Voltage (CC-CV) Charging Profile.2. Constant Current Charging Profile.Data was gathered by running simulations of a physics-based model. Achievements:1. For CC charging profiles, SoH estimation model built based on SoC data was efficiently able to estimate SoH based on data from different charging ranges. 2. For CC-CV profiles, models built on simulated data showed promising results on NASA's experimental data from batteries at different condition. 3. Proposed a modeling approach based on simulated data that could mirror real-life conditions.