Skip to content
View kaivalya-cyber's full-sized avatar
4 followers · 1 following

Achievements

Achievement: QuickdrawAchievement: YOLOAchievement: Pull Shark

Achievements

Achievement: QuickdrawAchievement: YOLOAchievement: Pull Shark

Block or report kaivalya-cyber

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
kaivalya-cyber/README.md
Kaivalya Singh
Typing SVG

LinkedIn Instagram Portfolio Email

Profile Views

whoami

const kaivalya = {
  school     : "Evergreen Valley High School  ·  Class of 2028",
  orgs       : ["Founder @ Synthica (RL/ML Research)", "Founder @ Vantage Point Learning"],
  focus      : ["Quantum Computing", "Reinforcement Learning", "Computer Vision"],
  stack      : ["PyTorch", "PennyLane", "MuJoCo", "Gymnasium", "OpenCV"],
  building   : "Adaptive Variational QEC Decoder  +  MARL Drone Swarm Research",
};

I work at the intersection of quantum computing, reinforcement learning, and computer vision — shipping real research and production-grade systems, not toy demos. I lead Synthica, a student-run international research org, and founded Vantage Point Learning, a STEM nonprofit serving 400+ students across five schools.

🔬 Research

  • Adaptive Variational QEC Decoder — CNN-routed syndrome decoder with 18.4% avg LER improvement, 27.2% peak
  • Reward shaping for nonlinear control — novel Lyapunov Satisfaction Rate (LSR) metric, PPO/SAC/LQR baselines
  • MARL Drone Swarm — CTDE architecture with emergent role specialization across two competing teams

⚙️ Engineering

  • PureGrad — autograd + backprop engine from scratch in NumPy, 99.7% acc, 19/19 tests
  • Drone Visual Tracking — YOLOv8 + SAC agent deployed on real drone hardware via MAVSDK/PX4
  • FTC Analytics Dataset — 53 events, 1,762 matches, OPR/ELO metrics, Kaggle published (88.69% acc)

tech_stack

PyTorch Gymnasium Stable--Baselines3 OpenCV YOLOv8 MediaPipe

PennyLane stim pymatching

MuJoCo PyBullet MAVSDK

Python C++ CUDA Java

React TypeScript Supabase Node.js

featured

⚛️ Adaptive Variational QEC Decoder
Real-time noise classifier routing syndrome data to specialized variational quantum decoders. 18.4% avg LER improvement, 94.2% classifier accuracy.
PennyLane PyTorch stim pymatching
→ repo

🤖 MAPPO Drone Swarm (MARL)
Centralized Training Decentralized Execution swarm — two competing teams of three drones with emergent role specialization.
PyTorch PyBullet Gymnasium
→ repo

🧠 PureGrad — Autograd from Scratch
Full deep learning framework in pure NumPy. Dynamic DAG, reverse-mode autodiff, MLP, optimizers. 99.7% acc, 19/19 tests.
Python NumPy Autodiff

🌐 Vantage Point Learning
STEM nonprofit platform with real-time volunteer feed, atomic claim logic, Google OAuth/PKCE. Serving 400+ students across San Jose & Asia.
React Supabase TypeScript
→ live · → repo

github_stats

"Sometimes the best way to solve a problem is to help others." — Uncle Iroh


Pinned Loading

  1. car_neural_diff_learning car_neural_diff_learning Public

    Python

  2. puregrad puregrad Public

    Python 1

  3. ftc-analytics-dataset ftc-analytics-dataset Public

    A clean, structured dataset of FIRST Tech Challenge (FTC) match results (2018-19 through 2023-24), with OPR metrics and ML benchmarks for win prediction and alliance strength.

    Python

  4. opponent-modeling-marl opponent-modeling-marl Public

    Multi-agent reinforcement learning with opponent modeling: PPO agents with GRU/Transformer opponent models, curriculum learning, past-self evaluation, and comprehensive analysis tools

    Python

  5. variational-qec-decoder variational-qec-decoder Public

    Python