Jun 2024 – Dec 2024 · 6 mos · Chennai, Tamil Nadu, India · Hybrid

• A Retrieval-Augmented Generation (RAG) chatbot was designed and developed using OpenAI/Claude APIs and LangGraph, enabling precise information retrieval from user documentation and internal knowledge bases.
• The chatbot’s reliability was enhanced through the implementation of strong guardrails, refinement of intent classification, and optimization of overall model performance. High-quality outputs were ensured by integrating comprehensive observability with Langfuse for interaction tracing and Promptfoo for LLM response validation.
• Structured logging and real-time alerting were incorporated via Sentry, enabling rapid detection and resolution of failures and unhandled exceptions.
• A Command Line Interface (CLI) tool was built for Aero-AI, an internal GenAI-based service. Using Node.js and the clack/prompts library, AI workflow setup and configuration management were automated, standardizing the handling of YAML files for system prompts, user prompts, and output schemas.
• Key commands—init for workflow setup, lint for YAML validation, compile for YAML-to-JSON conversion, and sync for updating workflows in DynamoDB—were implemented to ensure smooth and consistent integration with Aero-AI services.

Aug 2023 – May 2024 · 9 mos · Chennai, Tamil Nadu, India · Hybrid

• Developed a multi-step classification approach for unlabeled object views (wrl/stl files) using clustering algorithms (KMeans, GMM, HDBScan) and evaluated their effectiveness based on geometric properties.
• Explored deep learning architectures like Voxel CNN and autoencoders to extract advanced features for improved classification.

Jun 2023 – May 2024 · 11 mos · Chennai, Tamil Nadu, India · On-site

• Effectively managed diverse company interactions and provided insightful feedback to students on their resumes, while conducting interactive industry sessions to enhance students' career insights and job market preparedness.

May 2023 – Jul 2023 · 2 mos · Chennai, Tamil Nadu, India · On-site

• A variety of preprocessing techniques and machine learning algorithms were applied to optimize model performance, resulting in an F1-score of 0.73 with strong emphasis on both F1-score and recall. An interactive Streamlit dashboard was developed to visualize data drift and present estimated model performance metrics using the NannyML open-source library. Effective post-deployment model monitoring strategies were implemented to track model health and identify the ideal retraining intervals, ensuring sustained accuracy and reliability in production.

Jun 2022 – Jun 2022 · 0 mo · Mumbai, Maharashtra, India

May 2020 – Apr 2023 · 2 yrs 11 mos

May 2019 – Jun 2019 · 1 mo

• The infinitude of primes, the notion of congruence, the Quadratic reciprocity, the Quadratic
• Gauss sums, etc.
• The reference book: A Classical Introduction to Modern Number Theory by K. Ireland and M. Rosen.

Dec 2018 – Dec 2018 · 0 mo

• The project involved understanding the basics of Real Analysis.

May 2018 – Jul 2018 · 2 mos · Sahibzada Ajit Singh Nagar, Punjab, India

• Learned to use Gaussian 09, Gauss view, AIM2000, GAMESS.
• Performed UV-vis Spectroscopy, IR Spectroscopy, and ATR spectroscopy.