Jun 2025 – Aug 2025 · 2 mos · Seattle, Washington, United States · On-site

• Customer Support Voice AI Agents, Webex Contact Center Team
• Led a team of 4 interns to design voice AI agents, deployed pilot, scaling to 1.5M annual support calls
• Integrated agents with internal APIs and databases, reducing average call duration by 75%
• Built E2E testing/observability pipelines collaborating with Cekura (YC startup) accelerating deployment
• Developed RAG agents with LangGraph, scraping and indexing 500+ webpages to automate troubleshooting

Aug 2023 – Jun 2024 · 10 mos

• Undergraduate Thesis: Treatment Effect Estimation in Causal Inference
• Guide: Prof. Sunita Sarawagi, Computer Science, IIT Bombay
• Created a model-agnostic Individual Treatment Effect (ITE) estimator in JAX by developing a novel pair loss using approximate counterfactuals from the training data, reducing PEHE error by ~15% over state-of-the-art methods. Results were statistically significant across multiple datasets and model architectures. Published at ICML ’24.
• Developed the first tractable approximation for counterfactual representations from post-treatment covariates, enabling causal inference in settings where true counterfactuals are unavailable. I utilized counterfactual simulators to extract pre-treatment representations without access to true counterfactuals.

Aug 2023 – Dec 2023 · 4 mos

• English Language Training: Helped 40+ students build conversation skills empowering them to speak confidently through weekly classes

May 2023 – May 2024 · 1 yr

• Mentored 20 sophomores; advising on courses, projects, and internships

May 2022 – Feb 2023 · 9 mos

• Estimating Joint Probability Distribution With Low-Rank Tensor Decomposition, Radon Transforms and Dictionaries (Accepted at EUSIPCO 2023)
• Guides: Prof. Ajit Rajwade, IIT Bombay & Karthik Gurumoorthy, ML Scientist, Walmart
• Developed and implemented a joint density estimator using radon transforms of 2D marginals along with dictionaries
• Imposed a low-rank tensor structure using the Canonical Polyadic Decomposition of the joint density to reduce sample complexity for estimation
• Proposed a novel scheme for computing radon projections of continuous 2-dimensional marginals that reduced the problem to 1-dimension, achieving lower sample complexity over previous estimators
• Obtained lower Jensen-Shannon divergence compared to Gaussian Mixture Models and other methods

Jan 2022 – May 2022 · 4 mos

• MA105 Linear Algebra, TA: Conducted weekly tutorials for 50+ first-years where we discussed problem sets

May 2023 – Jul 2023 · 2 mos · Aalborg, North Jutland, Denmark · On-site

• Client Selection for Efficient Federated Learning, Advisors: Prof. Petar Popovski, Prof. Shashi Raj Pandey
• Developed a client selection strategy for communication-efficient federated learning, achieving faster convergence under heterogeneous data, computing resources, and privacy constraints. Published at IEEE Networking Letters.
• Improved training stability, reducing straggler impact by 90% under timing constraints and client heterogeneity
• Integrated efficient truncated Monte-Carlo sampling making combinatorial Shapley-Value approximation tractable

May 2021 – May 2022 · 1 yr · Mumbai, Maharashtra, India

• Presented before audiences of 100+ students on topics spanning microcontrollers, robotics, control theory
• Learned fundamental tools for robotics and electronics: ROS, Gazebo, OpenCV, AutoCAD, Arduino
• Ideated technical events for the institute community and looked after the overall functioning of the club
• Mentored several first-year students in projects, helping them debug hardware and software problems