Professional Experience

Ai & Perception Engineer

Feb’25 - Present

Designed and deployed real-time vision + control pipelines for UAV tracking, autonomous landing, and perception-guided navigation under strict edge-compute constraints.
Built low-latency video streaming systems (GStreamer + RTSP/UDP), sustaining 27–28 FPS from a 30 FPS camera feed on embedded platforms.
Implemented multi-core and multi-threaded optimizations in Python/C++ to reduce perception-to-actuation latency by 30%.
Worked across Jetson (Xavier/Orin), Raspberry Pi, and custom camera modules, contributing to sensor integration, camera evaluation, and model optimization for RGB + thermal pipelines.
Currently exploring GNSS-denied autonomy, focusing on early-stage research into vision-based navigation for UAVs.

Drawing Jan’23 - May’24
AI4CE Lab, New York University

Supervisor: Dr. Chen Feng, NYU

Developed and prototyped video-based VPR pipelines using CNN encoders, NetVLAD descriptors, and sequential matching for GPS-denied navigation.
Implemented multiple VPR algorithms (NetVLAD, SeqMatchNet, SeqNet) and built tooling for sequence construction, temporal feature aggregation, and distribution-based similarity comparisons (KL / JS divergence).
Designed experiments to evaluate robustness across sequence lengths, key-frame weighting, temporal ordering, and similarity metrics, leading to improved Recall@5 through optimized feature weighting strategies. Built end-to-end evaluation pipelines: KD-Tree ground-truth generation, sliding-window sequence construction, NetVLAD embedding extraction, and Recall@N benchmarking.
Worked with researchers to explore new ideas for distribution-based sequence descriptors, temporal invariance, and connections between image-VPR and video-VPR models.

Supervisor: Dr. Pushpak Jagtap, IISc, Bangalore

Led the setup and calibration of PhaseSpace Motion Capture System and developed ROS support for integrating the system with Turtlebot3 robots
Deployed Turtlebot3 robots with differential and mecanum drive configurations and utilized motion capture for accurate robot position tracking and real-time control adjustments
Implemented a Control Lyapunov Function (CLF) based controller in Python to guide unicycle modeled agents (robots) to desired poses
Implemented a barrier certificate-based collision avoidance algorithm for multi-agent systems using ROS(Python)
Achieved a 30cm safety radius in a 6m x 5m arena

Supervisor: Dr. Abhra Roy Chowdhury, Indian Institute of Science

Built an autonomous differential-drive rover using DC motors, encoders, a microcontroller, and a forklift-style mechanism for object manipulation.
Developed MATLAB-based models and control programs for autonomous navigation, including:
- Characterizing motor response through PWM speed sweeps and data analysis
- Implementing differential-drive kinematics for locomotion and arm control
- Deriving and simulating kinematic and dynamic models for open-loop motion control
- Designing and tuning a PID controller with encoder feedback for accurate position and heading control
Added a webcam and localization algorithms to estimate rover position and enable vision-guided navigation.
Integrated motion control, perception, and planning to perform high-level arena navigation and object manipulation tasks.
Enhanced autonomy with line-following and obstacle-avoidance behaviors.
Demonstrated strong skills in mechatronics, robotics modeling, control theory, computer vision, and autonomous decision-making.

M76 Analytics, Mumbai Aug’20 – Oct’20
Supervisors: Jai Mrug, Srikanth Atkuri

Developed automated data processing functions in Python to clean and preprocess real-time industry data
Enabled more efficient analysis by engineering the data into structured formats by 25%
Built scripts for tasks including data formatting, handling missing values, outlier detection, feature encoding, and normalization
Deployed data pipeline at the backend of the organization’s decision support system “MEGO”
Demonstrated skills in Python programming, data manipulation, pipeline development, and integration with business systems