Shreeyam Kacker

Shreeyam Kacker

Planet Labs · MIT · shreeyam [at] mit [dot] edu

Scholar · GitHub · LinkedIn · ORCID

Hey, I'm Shreeyam. I am a machine learning researcher working on representation learning and planning for autonomous systems. I received my PhD from MIT's STAR Lab, with PI Prof. Kerri Cahoy, where my research focused on autonomous tasking formulations for Earth-observing satellites using computer vision, onboard planning, and edge ML.

At Planet, I work on onboard ML and edge compute and autonomy for high resolution Earth-observing satellites. My previous experience also includes photonics at X, the moonshot factory (formerly Google [x]).

Education

Massachusetts Institute of Technology (MIT)

PhD Spacecraft Systems and Sensors, minor in AI and Computer Vision

2024 - 2025

Thesis: Spacecraft Autonomy through Computer Vision and Onboard Planning

SM Aeronautics and Astronautics

2020 - 2022

GPA: 5.0/5.0

Imperial College London

2016 - 2020

MEng Aeronautical Engineering with a Year Abroad

2016 - 2020

Degree Classification: First Class Honours

Experience

Planet Labs

2022 - 2024, 2025 - Present

ML Research Engineer

San Francisco, CA
  • Fine-tuned DINOv2 vision transformer (ViT) foundation model on Planet satellite imagery for scene classification and change detection
  • Distilled CLIP-style vision-language models to 8M parameters for lightweight onboard semantic filtering of satellite data
  • Designed and implemented algorithms for onboard dynamic task rescheduling using vision model outputs as inputs to a constraint-based planner
  • Deployed end-to-end inference pipeline for autonomous satellite operations, including model serving, camera drivers, and integration with the main flight computer
  • Architected next-generation onboard compute platform using NVIDIA GPUs as the first hire on the edge compute team
  • Radiation tested NVIDIA Jetson Orin and Thor at UC Davis cyclotron to qualify ML hardware for space operations; Thor results announced at NVIDIA GTC 2026

MIT Space Telecommunications, Astronomy and Radiation Laboratory

2020 - 2025

PhD Student (PI: Prof. Kerri Cahoy)

Cambridge, MA
  • Trained multi-agent reinforcement learning policies for electromagnetic formation flight, treating spacecraft as collaborative embodied agents performing continuous control in simulation
  • Developed reinforcement learning-based lookahead heuristics for dynamic task scheduling, outperforming greedy and analytic baselines on agile Earth-observing satellite constellations
  • Designed, trained, and validated deep learning models for cloud segmentation deployed onboard ESA's OPS-SAT satellite; validated on-orbit with real imagery
  • Created labeled dataset and trained edge detection models for ocean front detection from multi-spectral satellite imagery, published in IEEE TGRS

X Development LLC (formerly Google[x])

2021

Intern @X

Mountain View, CA
  • Interned under Taara, a free-space optical communications project; conducted architecture analysis and modeled integrated photonics for link budget optimization

Skills

ML Frameworks & Tools: PyTorch, JAX, CUDA, NumPy, SciPy, Pandas, Git, GCP, Docker

Programming Languages: Python, C, C++

Methods: DL, RL, CV, transformers, LLMs, VLMs, foundation models

Awards

  • First recipient of the Planet PhD Fellowship (2024)
  • Best Paper Award in Small Satellites at AIAA SciTech Forum (2022)
  • Imperial College London Dean's List (2019)

Publications

Continuous Attitude Planning for Agile Earth-Observing Satellite Collection Scheduling using Graphs of Convex Sets

Shreeyam Kacker, Kerri Cahoy

in prep.

Multi-Agent Reinforcement Learning for Spacecraft Electromagnetic Formation Flight with Transformer Policies

Josef Biberstein, Shreeyam Kacker, Kerri Cahoy, Sertac Karaman

in prep.; targeting ICRA 2027

Gaussian KL-Matching Regularization for Self-Supervised Learning

Shreeyam Kacker, Miles Shepherd, Kerri Cahoy

in prep.

Laser-guided space interferometer

Leonid Pogorelyuk, Paul Serra, Shreeyam Kacker, Sophia Vlahakis, Nicholas Belsten, Gioia Rau, Kenneth G Carpenter, Laurent Pueyo, John D Monnier, Ewan S Douglas, others

SPIE Optical and Infrared Interferometry and Imaging VIII Vol.12183, 2022

Other Projects

I started this blog for technical sewing projects. There are countless blogs from the early to mid 2010s that have been invaluable to my sewing journey, and I hope to contribute to that body of knowledge. I might occasionally write about fashion, research, or other topics also.

I wrote this blog platform called Spoingo. I also host a tree-based productivity software called Treetrack.