See the world.
Turning images and noisy sensor data into useful structure: pose, position, objects and state.
I build intelligent systems that move from computer vision to real-world hardware.
I’m Khushi, an Electrical Engineering Dual Degree student at IIT Bombay working across computer vision, robotics, autonomy and embedded systems.
My favorite problems sit at the boundary between algorithms and reality: getting perception to hold up outside a notebook, integrating subsystems, debugging edge cases, testing in simulation, and making the full stack reliable on hardware.
That has taken me from 3D human pose estimation at Sony R&D in Japan to autonomous underwater vehicles, fault-tolerant quadrotors and embedded control systems.
Turning images and noisy sensor data into useful structure: pose, position, objects and state.
Localization, planning and control for machines that need to act under uncertainty.
Simulation, integration and deployment, because the last 20% is usually where engineering gets interesting.
A curated set of projects that best represent how I think: systems work, technical depth, measurable performance and a bias toward building.
Reproduced a research pipeline for 3D multi-person pose estimation, benchmarked accuracy on a new setup, and evaluated whether live inference was practical on a consumer GPU.
Worked on the software stack powering Matsya from Oct 2023 to Jan 2026, spanning simulation, ROS1→ROS2 migration, acoustics-based localization, competition testing and R&D leadership.
Developed a control law for a quadrotor that could detect a single-motor failure, remain controllable with three motors, navigate and land safely.
I like environments where the problem is fuzzy, the stack is real, and the feedback loop is fast: research teams, robotics teams and hands-on engineering work.
Full story on LinkedIn ↗Worked on research problems in multi-person 3D pose estimation, reproducing a pipeline and evaluating both accuracy and live inference feasibility.
Establishing DAMP for the first time in IEOR and Applied Geophysics while also mentoring incoming first-year students through their transition into IIT Bombay.
Worked across simulation, acoustics localization, ROS migration, system integration and competition testing for Matsya, growing from the software team into an R&D leadership role.
Part of a core managerial team coordinating a high-volume placement season, training interview coordinators and owning recruiter processes across multiple sectors.
Built active obstacle-avoidance software for an unmanned underwater vehicle using MOOS-IvP, ROS integration, SONAR data and Raspberry Pi deployment; collaborated with the Indian Navy research team.
Building a foundation across control, embedded systems, signal processing, computer vision and digital systems while working on institute technical teams and research projects.
An early academic foundation in science and problem-solving: received the KVPY Fellowship with AIR 495 in the SA stream and the National Talent Search Examination scholarship.
Engineering taught me to make complex systems easier to navigate. Mentorship gives me a chance to do the same for people by creating structure, context and a reliable place to ask for help.
Establishing the Department Academic Mentorship Program for the first time across IEOR and Applied Geophysics. I’m helping create the peer-support structure for sophomore-and-above students, from mentor coordination and academic guidance to useful sessions, resources and dependable points of support within the branches.
As an ISMP mentor, I support new entrants through the shift from home and school to life at IIT Bombay: academics, campus culture, early choices and the personal adjustment that comes with a new environment. The role is less about having every answer and more about staying connected, listening well and helping mentees reach the right resources when they need them.
The common thread is systems thinking: whether it’s a robot or a student community, I enjoy creating structures that make difficult things easier to understand, navigate and trust.
Smaller experiments and course projects that filled in the systems picture, from microcontrollers to financial time series.
ROS2, Gazebo, RViz2 and OpenCV for package identification, path planning and simulated warehouse logistics.
Timers, interrupts, SPI, UART, ADC, LCD interfacing and board-to-board communication in Assembly and Embedded-C.
Built discrete-time convolution tooling to transform mono audio into spatial sound using BRIRs.
Financial time-series modeling, derivative pricing and risk analysis using ARIMA, GARCH, Black-Scholes and Monte Carlo methods.
Nine-segment serpentine robot with servo actuation, object detection and obstacle avoidance; 9th of 91 teams.
Customer discovery across freight and bus operators; pivoted to an IoT driver-assistance concept around fleet efficiency and drowsiness detection.
KVPY Fellowship, SA stream.
Percentile in JEE Main among 1.1M candidates.
AMD AI Sprint Hackathon finalist from 10K+ candidates.
Eureka! Business Model Competition participant pool; cleared Zonals.
If you’re working on computer vision, robotics, autonomy, embedded systems, or just an interesting problem worth thinking about, I’d love to hear from you.