May 2024 – Feb 2026 · 1 yr 9 mos · Bengaluru South · Hybrid

• Designed and deployed a multi-agent AI system (NXpert) using LangGraph, incorporating planner, router, and verifier patterns to optimize enterprise knowledge retrieval. Built a context-aware next-query planning agent with parallel generation and grounding via Elasticsearch scoring, reducing hallucinations and improving user experience. Developed a high-accuracy query routing system (98%) to dynamically select between retrieval and direct LLM responses, alongside query correction and multilingual normalization pipelines that improved retrieval metrics by 20–30%. Additionally, built a hybrid recommendation engine for Nutanix Learning Academy, increasing user engagement by 40–50%.

Oct 2021 – May 2024 · 2 yrs 7 mos · Bengaluru · Hybrid

• Developed a transformer-based information retrieval system over SOC specification documents to generate context-aware test plan suggestions, significantly improving efficiency for test engineers. Designed and implemented unsupervised similarity modeling for JIRA tickets using clustering techniques, enabling partial automation of issue triaging and reducing turnaround time through effective grouping of semantically similar tasks.

Aug 2020 – Oct 2021 · 1 yr 2 mos · Gurugram · Remote

• Explored adversarial robustness in deep language models, demonstrating the trade-off between resistance to perturbations and sensitivity to genuine semantic changes, and identifying conditions leading to over-stable yet semantically insensitive classifiers. Built a graph-based recommendation system using item-item co-occurrence graphs and attention-based GNNs to generate subgraph embeddings for context-aware basket recommendations. Also developed time-series forecasting models to predict user card usage behavior, leveraging sequential patterns for improved personalization and demand prediction.