My name is Ishaan and I am a recent graduate from Northwestern University (NU) with an MS in Computer Science and a BS in Mechanical Engineering. I was born and raised in Hong Kong, and attended King George V School.
I am currently a Graduate Research Software Engineer at the Presciene Lab at Northwestern University, working on computer / operating systems related projects including database/HPC application optimization, floating point computation, and compiler construction. I am also working on a statistical reinforcement learning and control ROS2 package in C++ inspired by research in Ergodic Imitation and Koopman Operators. Previously, I have also acted as a Software Engineering Intern at Tesla with the Electrical Systems team working on developing testing frameworks in Python and debugging firmware issues in C, as well as a Technical Program Manager Intern at Tesla scaling manufacturing lines with the Battery Team.
I am currently looking for full-time opportunities in Software Engineering, Quant, and Robotics starting in Winter 2024.
Website: https://ikn1062.github.io/IshaanNarain/
Email: [email protected]
LinkedIn: https://www.linkedin.com/in/ishaan-k-narain/
Below, I have listed some of the projects I have worked on as part of my time at Northwestern University. These projects include some of the work as part of my coursework, personal projects, and research projects.
-
Ergodic Metric and Controller Library for Dynamic System
- Created C++ reinforcement learning library with model predictive control for autonomous robotics inspired by ergodic metric research
- Derived mathematics for iLQR control via gradient descent by solving Riccati equations from derivative of ergodic objective
- Engineered interactive 3D learning-from-demonstration cart-pole simulation using Linux Robot-OS 2, CMake, and C++
-
Zoom (and Live) Video Speaker Data Analysis
- Developed Zoom and Live (OWL) video data collection and analysis app, leveraging OpenCV, Mediapipe, and face recognition Python Libraries
- Utilized 30 face-mesh tracking points to analyze speaker facial cues to predict the speaker when masks are worn
- Reduced processing time by 82% using multithreading and multiprocessing functions, executing code in NU Linux computing cluster
-
DELOS - Deep Learning Model Serving System
- Implemented automated statistical stress-testing-modules for analyzing performance of re-trained Neural Networks
- Decreased training time by 24% by dynamically re-optimizing weights using online Multi-Arm Bandit memory replay algorithm
- Developed stateless full-stack app to improve monitoring of over 20 ML KPIs, leveraging TensorFlow, Mongo, and Kubernetes
- Some of my work in the application process is linked here
-
Sensing Navigation and ML in Robotics
- Created robotics library for Extended Kalman Filtered simultaneous localization and mapping using ROS2 and Armadillo in C++17
- Integrated DDS publish-subscribe network protocols to facilitate communication between sensors and robot actuators
- Automated packaging, documentation, and testing of 20,000-line C++ library using CMake, Doxygen, and Catch2 Framework
- A simulation demonstration video of SLAM is linked here
-
- Designed and built wearable haptic robot to simulate weight of objects from VR up to 10N by combining torques from 3 motors
- Derived kinematic model to estimate arm positions and calculate torque using data from Kalman-filtered sensors and Unity VR
- Programmed control loop, robot geometry, and serial communication to run in 0.001 seconds per loop with embedded C++
- โข Created and manufactured electrical board integrating 38 devices including motors, encoders, and 1500W power supply
- Haptic Robot for simulating the weight of objects in VR
-
Venmo Request Scheduler (NU WildHacks 2022 Winner)
- Launched full-stack Venmo recurring payments scheduler web app on team of 4 in 24 hours, winning NU22 Hackathon prize
- Created 20 unique API endpoints and logic on Flask server for back-end and front-end integration, recurring payments, and messaging process
- anually overrode class methods from Python Venmo API wrapper to repurpose for app login and 2-factor OTP authentication
-
Northwestern University Solar Car Team - Previous Project Manager and Lead Mechanical Engineer
- Managed team of 40 to design, manufacture, and assemble road-legal carbon-fiber solar vehicle to compete at Formula Sun Grand Prix
- Engineered new suspension and monocoque chassis, increasing total power-to-weight ratio by 13% by using mechanical calculations and Matlab Simulink
- Led team to design and build new powertrain of car integrating battery pack, battery management system, solar array, and motor, improving solar power generated by 8%, power to weight ratio by 13%, and accomplishing an indefinite range of vehicle at 30mph
- Coordinated with University administration, faculty, and sponsors to secure funding and material resources over $60,000 of value
- Generated 40-page report detailing and analyzing the performance of vehicle systems proving compliance with 300 racing regulations
- Refined driver and vehicle control communications by integrating 8 vehicle components using CAN protocol in embedded C++