Experience

  1. 01

    Jul 2026—Sep 2026

    Remote / Vellore

    Robotics Research Intern

    Department of Robot and Smart System Engineering, Kyungpook National University

    • Worked with Dr. Yuvaraj and Dr. Bubryur Kim on autonomous exploration and hazard mapping for robotic inspection in unknown environments.
    • Developed CRAFT, combining C*-based coverage planning with CUSUM gas anomaly detection. Suspected gas hotspots trigger finer local sweeps to improve concentration mapping and source localization.
    • Integrated visual hazard detection with depth measurements and robot pose, locating hazards on the map and updating navigation costs so the rover can avoid them.
    • Evaluated gas-triggered refinement in four GADEN simulation environments. Comparisons in three environments showed approximately 10–59% lower mean localization error than coverage without adaptive refinement.
    • Deployed the integrated ROS 2 system on an NVIDIA Jetson Orin Nano rover, running navigation, gas monitoring, and visual hazard mapping onboard, with visual processing at approximately 15 FPS during the indoor deployment.
  2. 02

    Jan 2026—Jul 2026

    Remote

    Machine Learning Research Intern

    University of Texas Rio Grande Valley, AIM Lab

    • Worked under Prof. Jinwoo An, collaborating with civil and chemical engineering researchers on materials prediction and crack assessment.
    • ECC prediction: developed CatBoost models to predict the tensile strength and ductility of engineered cementitious composites from their mix design. Grouped 620 tests into 276 unique mixes to prevent leakage between repeated tests of the same composition.
    • Uncertainty and inverse design: calibrated prediction intervals and connected the models to NSGA-II optimization for cost-efficient mix design, with constraints to keep recommendations within the supported data domain. Nominal 80% intervals achieved 80.4% stress and 82.6% strain coverage.
    • Crack measurement: built a replaceable vision front end that segments cracks, measures their paths, and uses MobileSAM to estimate member boundaries. Calibrated image measurements provide geometry for the physics backend, alongside supplied load and material properties.
    • Physics-informed fracture modeling: developed a mesh-free PINN with an embedded Williams crack-tip field to estimate Mode I stress intensity. Evaluated the mechanics solver separately from the vision pipeline, obtaining approximately 6.5% error against an independent finite-element reference in the five-seed reference experiment.
  3. 03

    Oct 2024—Jun 2025

    Remote

    PRISM Research Intern

    Samsung Research Bangalore

    • Selected through Samsung Research’s PRISM fellowship.
    • Engineered BQ2Viz, a multi-agent GenAI framework (LangGraph, LangChain, ChromaDB) that converts natural language into data visualizations, using a Profiling Agent for automated schema indexing and a Self-Correcting Code Generator.
    • Designed a Reasoning-Based Planner for multi-file retrieval, inferring multiple distinct visualizations from a single abstract query (e.g., “Analyze Sales” → trend line + category bar + regional pie).
    • Benchmarked against NVAGENT on the VisEval dataset: 84.57% accuracy on complex multi-table queries, cutting the error rate to 11.5% via a feedback-driven code-repair loop.
  4. 04

    Apr 2024—Sep 2025

    Vellore

    Autonomous Systems Engineer

    Team Vyadh (Mars Rover Team), VIT

    • Engineered the autonomous navigation stack with ROS 2 (Nav2) and SLAM, implementing a Frontier Exploration algorithm to map unknown environments within dynamic GPS geofences.
    • Built a real-time obstacle & pit avoidance perception pipeline using Intel RealSense D455 + PCL, with odometry-based path recovery and a custom PID alignment to compensate for wheel drift on uneven terrain.
    • Fused IMU + odometry with an Extended Kalman Filter to cut state-estimation noise, and built a telemetry GUI for Astro-Bio sensor streams (CO₂, IR, moisture).
    • Part of the team that placed 13th at the International Rover Challenge (IRC) and 4th at the IRDC 2025.

02 / Education

VIT Vellore

B.Tech. in Computer Science and Engineering

Sep 2022 – Jul 2026