2023 to present
IIT Tirupati
B.Tech, Mechanical Engineering
Mechatronics, controls and Theory of Machines, modelled hands-on in MATLAB, Simulink and Simscape.
01About
I'm Kanishk, a mechanical engineering student at IIT Tirupati who is happiest where code meets hardware. Alongside my B.Tech I'm working through the BS in Data Science and Applications at IIT Madras, which grounds my machine learning in statistics.
Most of what I build lives in simulation first: ROS 2 robots in Gazebo, arms planned with inverse kinematics in MATLAB, agents trained with reinforcement learning in PyBullet. The goal is always the same, autonomous systems where algorithms meet actuators.
01
I take a problem apart until every piece makes sense, then put it back together in code.
02
The interesting part is the loop: perception feeding decisions feeding actuators.
03
Mechanical precision matters to me as much as the flexibility of software.
04
Every simulator, solver and model I learn makes the next build faster.
2023 to present
B.Tech, Mechanical Engineering
Mechatronics, controls and Theory of Machines, modelled hands-on in MATLAB, Simulink and Simscape.
Online degree, ongoing
BS, Data Science and Applications
Machine learning, statistics and analytics in a Python-based curriculum.
02Projects
Fifteen builds from my GitHub, from ROS 2 simulations and inverse kinematics to reinforcement learning, computer vision and the web. Each one has a write-up and a link to its code.
Showing 15 projects
Robotics · Computer vision
A ROS 2 rover that maps its surroundings with lidar, watches the road with a camera and names what it sees with YOLOv8.
Computer vision · Machine learning · Robotics
Renders 3D models into randomised, auto-labelled datasets, trains YOLOv8 on them and deploys the detector as a ROS 2 node that fuses detections with lidar range.
Robotics · MATLAB & Simulink
A MATLAB motion pipeline that solves inverse kinematics for Cartesian waypoints with its own damped least-squares solver, then blends them into a C¹-continuous joint trajectory.
Robotics · Computer vision
A ROS 2 package that models a two-joint arm with a camera on its own link, drives it along a square trajectory in Gazebo Fortress and visualises joint states and TF in RViz2.
Machine learning · Robotics
A Gymnasium environment around a simulated KUKA iiwa, trained with Stable-Baselines3 PPO to bring its end-effector within 5 cm of a randomly placed target.
Machine learning
A multi-agent DQN trainer for Snake with a custom Gymnasium environment, a live Pygame panel for every agent, resumable runs, fine-tuning and a showcase of the best agent.
Robotics · Machine learning
A differential-drive simulator with five IR sensors trained by tabular Q-learning, then ported into a Webots world with downward cameras as the line sensors.
MATLAB & Simulink
Course project that models a shaper machine's quick return mechanism in MATLAB R2025b Simscape Multibody and studies the ram's position, velocity and acceleration.
MATLAB & Simulink · Robotics
MATLAB and Simulink models of a planar two-link arm that follows sine references in joint space and traces a square with its end-effector using analytical IK.
Machine learning
A complete scikit-learn pipeline that classifies tumours as benign or malignant from 30 cell-nucleus features, reaching about 97% test accuracy and 98% recall.
Web & tools · Robotics
Co-built with Atharva Bhosale: assemble a robot joint by joint, then explore DH parameters, kinematics, Jacobians, trajectories, dynamics and workspace with nothing to install.
Robotics
A PySimVerse drone that takes off, tunes its speed and alternates forward and sideways moves to approximate a diagonal route, with a keyboard teleop script alongside.
Computer vision
A Python detection script built on the YOLO11 medium model and OpenCV that draws labelled boxes with confidence scores, frame by frame.
Computer vision
Python scripts that read a PLY scan and render it with Open3D, PyVista and PyntCloud, plus voxel-grid downsampling with a tunable voxel size.
Web & tools
A no-backend web app that downloads comics from Google Drive, renders them with pdf.js, caches them in IndexedDB, remembers your page and groups series automatically.
Course project notebook for supervised learning, from preprocessing and feature engineering to tuned Random Forest and XGBoost models.
Jupyter Notebook · 2025
Python utility that tidies a folder by sorting its files, with the sorting logic kept separate from its desktop interface.
Python · 2026
Experimental workspace that pairs a Python app with a C++ module and a URDF robot model.
Python · 2026
An earlier portfolio site, hosted on GitHub Pages.
HTML · 2026
03Toolkit
Everything here shows up in at least one repository, from simulators and solvers to the libraries that train and see.
Primary language of each of the 24 public repositories, as GitHub detects it.
Other covers Simulink and Simscape models and README-only repositories, which have no detected language.
04Journey
From a first classifier to ROS 2 rovers and reinforcement learning: how the work has grown since I started engineering in 2023.
2023
Picked up Python and the basics of machine learning alongside core mechanical coursework.
2024
Built a logistic regression classifier for breast cancer diagnosis with a full evaluation pipeline, reaching about 97% test accuracy.
2025
Simulated a 2-DOF arm with trajectory tracking in Simulink, and modelled a six-link quick return mechanism in Simscape Multibody for Theory of Machines.
Jan to Mar 2026
Brought a URDF arm up in RViz2 and Gazebo, then gave it a live camera and a square trajectory. Ran first real-time detection experiments with YOLO11.
Apr to May 2026
Built RoverX with lidar, a camera and YOLOv8, flew scripted routes in PySimVerse, worked with point clouds in Open3D and wrote a damped least-squares IK solver for the Kinova Gen3.
Jun 2026
Trained a KUKA iiwa to reach targets with PPO in PyBullet, and ran up to twelve DQN agents side by side on Snake.
Aug to Sep 2026
Ported a Q-learning line follower into Webots, built a pipeline that trains YOLOv8 from rendered 3D meshes, and shipped Gutter, an offline-ready comic reader.
Oct 2026
Co-building a browser-based robotics analysis studio with Atharva Bhosale.
Next
Sensor fusion and closed-loop feedback on the ROS 2 arm, a real-time YOLO and OpenCV pipeline, deep learning and C++.
05Contact
Open to ML internships, robotics R&D and research collaborations in 2026. Send a message here, or reach me directly.
GitHub
@KanishkPatel0206