Thanks to visit codestin.com
Credit goes to github.com

Skip to content
View nightcrab's full-sized avatar
8 followers · 8 following

Achievements

Achievement: Pair ExtraordinaireAchievement: Pull Shark

Achievements

Achievement: Pair ExtraordinaireAchievement: Pull Shark

Block or report nightcrab

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.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

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

Report abuse
nightcrab/README.md

Paul Anderson

Parallel computing・Machine Learning・Game Development

Python C++ MCTS Deep RL Tetris AI

Personal Site · LinkedIn · ANU · Kyoto University


I'm interested in high-performance search algorithms, parallel tree search, self-play reinforcement learning, and making strong AIs for imperfect information two-player games — especially falling-block games.

I'm also the developer of the live-service competitive block stacker, God of Stackers. Interested in the game? Join our Discord server!

Currently most active on:

  • Parallel search algorithms with advanced load-balancing
  • Statistical, Monte Carlo and ML approaches to problems in game theory
  • Project managing and developing God of Stackers

🌟 Current Projects

Nana

Parallel MCTS engine for versus Tetris / tetromino battlers

  • Parallel Monte-Carlo Tree Search with Transposition Driven Scheduling (TDS) load balancing
  • Bitboard game state + fast move generation
  • Custom message-passing architecture for scalability
  • Online any-time planning

God of Stackers

Live service, competitive multiplayer stacker with regular updates

  • Real time, online multiplayer using WebSockets
  • Scaling backend using multiple cores + Redis pub/sub
  • High perfomance client with lower input latency than both TETR.IO and jstris
  • Daily active users and >1,000 registered players
  • Major updates every month + minor patch every week

🌟 Previous Work / Contributions

Ivy

Open source interoperability between popular machine learning frameworks

  • Major contributor and early engineer (2022-2024)
  • Enabled model and training pipeline conversion from Torch to JAX and Tensorflow
  • Performance improvements across CNN, RNN and Transformer architectures
  • Computation graph approach ensured robust and correct transpilations

Pinned Loading

  1. nana nana Public

    Tetromino-battler AI using parallel MCTS, bitboards and message passing architecture.

    C++ 3

  2. ivy-llc/ivy ivy-llc/ivy Public

    Convert Machine Learning Code Between Frameworks

    Python 14.2k 5.5k

  3. apollyon-rag apollyon-rag Public

    High quality Retrieval Augmented Generation (RAG) with a local LLM + web interface.

    Python 1

  4. nb-stream-compaction nb-stream-compaction Public

    In-Kernal stream compaction written in Numba CUDA for efficient general purpose GPU algorithms.

    Python 1

  5. contests contests Public

    Solutions for contest problems (CodeForces, USACO, AIO).

    C++

  6. tromis-net tromis-net Public

    Playing Tromis with deep reinforcement learning. C++ client, pytorch NNs, and replays.

    Python