Abhishek M.
Chandure
Hi, I'm Abhi — an engineer, software developer, and systems thinker driven by curiosity and continuous growth. I work at the intersection of aerospace, software, AI, and intelligent systems, with a passion for solving complex real-world problems through technology.
I design and build systems with precision and purpose. The same systems-thinking mindset used in aerospace — understanding dynamics, control, and interconnected behavior — now drives my journey into AI and machine learning. Because the next frontier isn't just building systems that observe the world; it's building systems that can understand it.
Beyond technology, I'm deeply invested in self-improvement, fitness, and long-term thinking. I believe meaningful progress comes from disciplined habits, learning relentlessly, and constantly pushing beyond comfort zones.
This space is where I share my projects, ideas, experiments, and my journey of building better systems — and becoming better in the process.
What I actually work on.
A Kalman filter taking readings from magnetometers, sun sensors, gyroscopes and an IMU, fusing them into a best-guess attitude, and commanding reaction wheels to keep the satellite pointed at a patch of Earth. All of it in C++ and MATLAB-Simulink, running for years unattended.
PyQt Plotting Prototype — Desktop Visualization Framework
Designed and developed a desktop-based high-performance plotting prototype using Python and PyQt, replacing browser-based visualization for engineering workflows. Redesigned the plot_cd_mach_graph() function to accept figures and dataframes dynamically, added support for multidimensional arrays, and unified multiple graphing requirements into a single modular framework. Delivered iterative demonstrations to engineering teams, implemented stakeholder feedback, and successfully completed requirements 1–4, 9–12.
DANCE Plotter Architecture Study & Integration
Studied the DANCE Plotter architecture to align the visualization prototype with existing engineering standards. Mapped new implementation against framework requirements, ensured compatibility between prototype components and architecture design, and worked closely with engineering teams to maintain long-term maintainability and scalability.
CMAKE Library Build & Integration
Configured build pipelines, built and integrated engineering libraries using CMAKE, and troubleshot dependency issues for internal teams. Assisted domestic teams in library development and delivered all required libraries on schedule. Received appreciation from project stakeholders and the product manager for reduced integration effort and improved build stability.
Earth-Observation Cube Satellite
Designed cube satellites for 5-year nominal missions in low-inclination LEO (500–600 km altitude), delivering hyperspectral and multispectral imagery at ~3 m resolution for agriculture, defense, energy, and banking. Owned the Attitude Determination & Control System end to end — magnetometers, sun sensors, gyroscopes, reaction wheels and IMU, fused through a Kalman filter, with the control loop running in C++ and MATLAB-Simulink. Also led collaboration outreach with national and international firms across multiple satellite programs.
Heavy-Payload Drones with High Endurance
Designed a methodology for multicopters and multirotor UAVs carrying heavy payloads over long ranges and flight durations. Built a Python algorithm that searches the combinatorial space of motors, propellers, batteries, and frames against real load-test data to surface the optimal configuration for a given payload and endurance target. Modelled and simulated quadcopter and hexacopter variants in Python and Matlab-Simulink, with stabilization and trajectory control via Euler-angle PID.
Control Laws for UAVs Under High Wind
Literature review and comparative study of wind-field estimation using a standard autopilot sensor suite, and controllers for rejecting wind disturbance. Examined Active Disturbance Rejection Control (ADRC) against PID, and Linear Quadratic Regulator with Integrator (LQRI) against LQR. ADRC outperformed PID on disturbance rejection; LQRI outperformed LQR under high wind.
Launch Simulation — Long March 3B (Apstar-6C)
Full end-to-end launch-vehicle simulation in Python. Gathered mission data — orbital parameters, trajectory, launch configuration — and designed a nominal trajectory that satisfied the required terminal conditions. Clean, standalone simulation with all components in one codebase.
Rocker-Bogie Rover with Self-Balancing Platform
Designed and built a six-wheeled rover with rocker-bogie suspension and an IMU-stabilized platform on top — so the payload stays horizontal regardless of terrain. CAD in Solidworks, kinematic simulation in Adams, C++ firmware on Arduino UNO for both the drive system and the platform controller.
Thermoacoustic Refrigeration
Built an experimental thermoacoustic refrigeration setup from scratch. Studied multiple stack materials and positions, took temperature readings across configurations, plotted and compared the results to draw conclusions about performance. One of my first hands-on lessons that elegant physics and messy hardware are inseparable.
Transition into AI & Data Engineering
Working through a structured path covering statistics, classical ML, deep learning, data science in Python and R, NLP, and Spark/Scala. End-goal: ship end-to-end projects that pair my existing controls and signals background with modern ML — especially around satellite imagery, remote sensing, and time-series data. This section will grow as I publish.
Meditations
Thinking, Fast and Slow
Man's Search for Meaning
Zen and the Art of Motorcycle Maintenance
The Beginning of Infinity
Range
A working notebook — placeholder prompts in the spirit of the astrophysics, controls, and cosmology threads I care about. Rewrite, cut, or expand any of them into full essays when the thought finishes forming.
Why does the Kalman filter feel like a miniature theory of perception?
Predict, then correct with noisy evidence, then predict again — weighted by how much you trust the model versus the sensor. It's a recursive procedure for being less wrong over time. What does it mean that our best estimator for a spacecraft's attitude is also a decent cartoon of how a mind updates beliefs?
The Fermi paradox: is the great filter behind us, or ahead?
If the galaxy should be teeming with civilizations and isn't, then something systematically stops the jump from single cell to starfaring species. Either we already made the hard crossing (rare abiogenesis, rare eukaryote step), or we haven't yet. The question isn't academic — it changes what kind of century this is.
A 3U cubesat at 550 km — what is it actually fighting?
Residual atmospheric drag, solar radiation pressure, magnetic torques from Earth's field, gravity gradient, reaction-wheel saturation, sensor noise. None of them dramatic individually. All of them at once, for five years, with no maintenance. Orbital lifetime is a slow-motion argument against entropy.
What would it take to directly image an Earth-like exoplanet?
The star is roughly ten billion times brighter than the planet. You need coronagraphs, starshades, or nulling interferometry operating at angular resolutions that make the Hubble look myopic. And then you need to hold that geometry stable for hours. It's an ADCS problem dressed up as cosmology.
Treating a career like a trajectory problem.
A trajectory has initial conditions, terminal constraints, control authority, and noise. Most people optimize step by step and end up nowhere in particular. Pick the terminal condition first — the orbit you want to be in — then design the control law. Recompute when the state diverges. Don't confuse turbulence with drift.
Consistent inputs over long time horizons.
A satellite's lifetime isn't decided by any single maneuver — it's decided by thousands of small, accurate corrections over years. Training is the same problem in a different coordinate frame: the work is mostly boring, mostly invisible, and mostly done on the days you don't feel like it.
Five days a week of strength work. Daily cardio. Sleep is the real training tool. Currently on a slow body-recomposition run, targeting ~15% body fat while holding strength numbers.
The parts of the day that don't fit in a commit message — books, lifts, weekend walks, plates of food, the odd sunset. A slice of the life behind the work.
follow on instagram / @abhichandure01 →photos/ folder next to index.html, then update the tiles below with filenames like photos/lift-day.jpg. For live Instagram sync, add an EmbedSocial or LightWidget embed (~$5/mo) into this section. Also swap @your_handle in the link above.
