Robotics Engineer & Researcher

Project Overview

Developed an ROV-based system for retrieving abandoned "ghost" lobster pots from the ocean floor. Ghost lobster pots pose a significant threat to marine ecosystems since they continue to trap and kill marine life long after abandonment.

Key Contributions

• Architected localization stack integrating USBL, DVL, and GPS for 3m live 3D positioning using Python
• Characterized USBL accuracy through pool testing; validated navigation pipeline
• Conducted integrated tests in pool and Charles River, demonstrating pot retrieval with Ping 360 sonar and camera
• Designed and tested custom spear mechanism for engaging and retrieving lobster pots

Technical Details

The system integrated GPS-tagged sonar data, a USBL/DVL-based localization module, and autonomous waypoint navigation via QGroundControl to locate and move to the pot location. Once within six to eight meters of a tagged pot, the ROV used onboard sonar and cameras to identify and latch onto the pot using a custom spear mechanism.

Results

Testing in the Zesiger Center pool and Charles River demonstrated the system's ability to reliably locate and retrieve pots with improved accuracy and reduced effort compared to traditional methods. This approach offers a scalable, cost-effective solution for mitigating ghost gear impacts and supporting sustainable lobster fisheries.

Project Overview

Developed a finite-dimensional, control-affine model of a cable-driven snake-arm robot under the constant-curvature assumption, and designed a robust trajectory-tracking controller.

Key Contributions

• Modeled a continuum robot's motion to enable accurate real-time control of flexible multi-segment arms
• Developed and implemented a robust adaptive controller to maintain precise trajectory tracking under uncertainty and disturbances
• Translated desired joint movements into actionable actuator commands while ensuring safe tension limits for all cables
• Verified system stability and reliability through rigorous testing and analysis, ensuring predictable performance in real-world scenarios

Technical Details

Using the piecewise-constant-curvature (PCC) model, derived the forward kinematics via homogeneous transformations and computed the cable Jacobian that relates joint torques to cable tensions. Designed a sliding-mode controller with adaptive parameter estimation based on joint-space dynamics. At each control step, a quadratic program (QP) was solved to map the desired torques to physically feasible nonnegative cable tensions.

Results

Stability was analyzed using Lyapunov methods, showing exponential convergence of both tracking error and parameter estimates within the sliding tube. Contraction analysis was also employed to certify global robustness and convergence guarantees. The resulting control scheme achieved accurate trajectory tracking under bounded disturbances and model uncertainty, while respecting cable-actuation constraints.

Project Overview

Built an autonomous robot for the MIT MASLAB competition that could identify, collect, and stack colored cubes. The robot was designed to locate stacks of cubes, knock them over, identify cubes of the team's assigned color, and stack them into towers.

Key Contributions

• Built sensor fusion pipeline with IMU, webcam, ToF, and wheel odometry
• Programmed low-level ESP32 interfaces (C++) and high-level control on Jetson Nano (Python)
• Implemented sparse-graph path planner and precise timed motion profiles for autonomous stacking
• Co-designed mechanical subsystems (chassis, lift, actuated door) in Onshape

Technical Details

The robot used a four-step process: 1) Ram into a stack and knock over cubes, 2) Funnel cube into center tower, 3) Spear through cube's center hole with linear actuator, 4) Lift cube with rack and pinion mechanism. Once a cube was lifted, the process repeated to stack additional cubes. The robot would then lower the cube tower, unspear the bottom cube, open the front panel with a servo, and back away leaving a completed stack.

Results

Our team achieved 2nd place out of 12 teams in the competition. The robot successfully demonstrated autonomous cube identification, collection, and stacking capabilities within the competition timeframe.

Project Overview

Developed automated optical characterization workflow for specialty integrated circuits and contributed to research on high-parallelism silicon micro-ring modulators. My work focused on laser trimming techniques to compensate for fabrication errors in photonic devices.

Key Contributions

• Automated optical characterization workflow for specialty ICs, reducing test time from 5.3 hours to 40 minutes per 32 test sites (90% improvement)
• Designed 3D-printed optical mounts in Fusion 360, saving $200/unit and improving alignment repeatability
• Developed synchronized Python control scripts for linear stages and servos enabling unattended operation with <1% measurement error
• Implemented continuous acquisition methods for real-time resonance tracking during laser trimming
• Contributed to research on 64-channel silicon micro-ring modulators with compact footprint, high extinction ratio, and large bandwidth

Technical Details

The project involved developing methods for holographic annealing (laser trimming) of photonic integrated circuits. This included creating both initial methods using start-stop resonance measurements and updated methods using continuous monitoring of resonance changes during trimming. The system used a 405nm laser diode for trimming and various measurement lasers (780nm velocity laser, M² systems) for resonance characterization.

My work on laser trimming contributed to the larger research on 64-channel parallel ring modulators, which demonstrated dense integrated micro-ring resonators with non-volatile laser trimming to compensate for fabrication errors.

Results

Successfully demonstrated non-volatile laser trimming of silicon micro-ring modulators to compensate for fabrication errors. Achieved precise resonance shifting with the ability to actively target specific resonance wavelengths. The research on 64-channel modulators was published and presented at CLEO 2025, showing compact footprint (4-μm radius), high extinction ratio (>20 dB) and large bandwidth (>25 GHz).

Project Overview

Designed and built a large LED display with 1500 individually addressable LEDs controlled via the FastLED library. Managed the project from concept to completion with a team, including securing funding and presenting at an expo.

Key Contributions

• Applied for and secured $500 in funding from ProjX for materials
• Designed the layout and electrical system for 1500 LEDs on a wooden panel
• Soldered all connections and implemented control using the FastLED library
• Coordinated team efforts and presented the completed project at the ProjX Expo

Technical Details

The project involved attaching 1500 LEDs to wooden paneling, with careful planning of the electrical layout to ensure even power distribution. Copper sheets were used for power buses, and all LEDs were wired together in a grid pattern. The FastLED library was used to control the LEDs, allowing for complex lighting patterns and animations.

Results

Successfully completed the LED board within the winter session timeframe and presented it at the ProjX Expo alongside other participants. The project demonstrated skills in electronics, embedded programming, and project management.