Skip to content
View zapabob's full-sized avatar
🎯
Focusing
🎯
Focusing

Organizations

@happyscienceiscult

Block or report zapabob

Report abuse

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

Report abuse
zapabob/README.md
zapabob AI engineering portfolio banner

AI Engineering Portfolio // zapabob

AI engineer building drug-discovery models, agent systems, CUDA measurement experiments, and VR/Unity research tools.

Selected AI Engineering Results

Problem AI engineering work Evidence Repo
Drug-discovery signal modeling Multi-target pIC50 platform for psychoactive and drug-like compounds using Transformer regression, RDKit descriptors, SMARTS scaffolds, and Optuna DAT, 5HT2A, CB1, CB2, and opioid receptor workflows with GUI/CLI and session recovery multi-target-pIC50-predictor
GPU/LLM systems measurement KV-cache TurboQuant Pareto experiments for Qwen3.5-9B on CUDA with captured replay Measurement-focused repository for reproducible LLM systems experiments Turboquant-CUDA
Experimental model research Elastic Looped Transformer causal-LM port with Intra-Loop Self-Distillation and GRPO PyTorch research implementation for advanced architecture experiments elastic-looped-transformer
zapabob AI engineering portfolio dashboard

Portfolio Repositories

Repo Why it matters
multi-target-pIC50-predictor AI drug-discovery workflow: descriptors, Transformer regression, target-specific pIC50 prediction, GUI/CLI
Turboquant-CUDA CUDA and LLM systems measurements for KV-cache TurboQuant Pareto analysis
elastic-looped-transformer Experimental PyTorch model research: ELT, self-distillation, GRPO
jaxa-earth-vrchat-terrain Data pipeline from JAXA earth-observation inputs to Blender, Unity, and VRChat terrain
hermes-agent Agent tooling and local AI assistant experimentation
liltoon-pcss-extension Unity/VRChat C# extension work for production-style shader tooling

Engineering Focus

  • AI for drug discovery: pIC50 prediction, cheminformatics, RDKit-style descriptors, receptor-target workflows
  • LLM systems: CUDA measurement, KV-cache experimentation, model-efficiency tradeoff analysis
  • Agentic software: autonomous assistant tooling, workflow automation, local-first research systems
  • Research implementation: PyTorch experiments, architecture ports, reproducible exploratory code
  • Spatial computing: Unity, Blender, VRChat, terrain pipelines, and creator tooling

Connect

Snapshot Notes

  • Profile snapshot date: 2026-05-26 Asia/Tokyo
  • Public profile counts and repository telemetry were refreshed from GitHub REST API and the public profile page
  • Current public surface: 166 public repositories, 68 owned public non-fork repositories, 231 followers
  • Current public pinned order: multi-target-pIC50-predictor, jaxa-earth-vrchat-terrain, hypura, Turboquant-CUDA, elastic-looped-transformer, hermes-agent

Pinned Loading

  1. multi-target-pIC50-predictor multi-target-pIC50-predictor Public

    Advanced multi-target pIC50 prediction platform for psychoactive and drug-like compounds. Supports DAT, 5HT2A, CB1, CB2, and opioid receptors. Features Transformer regression, RDKit descriptors, SM…

    Python 1

  2. jaxa-earth-vrchat-terrain jaxa-earth-vrchat-terrain Public

    Transform JAXA satellite data into VRChat worlds! Generate 3D terrain from Earth observation data for Blender/Unity/VRChat. Works with Cursor/Codex IDE.

    Python 3

  3. hypura hypura Public

    Forked from t8/hypura

    Run models too big for your Windows memory and GPU

    Rust 2

  4. Turboquant-CUDA Turboquant-CUDA Public

    Qwen3.5-9B KV-cache TurboQuant Pareto measurements on CUDA with captured replay

    Python 7

  5. elastic-looped-transformer elastic-looped-transformer Public

    ELT (Elastic Looped Transformer, arXiv:2604.09168) + Intra-Loop Self-Distillation + GRPO, causal-LM port in PyTorch

    Python 3

  6. hermes-agent hermes-agent Public

    Forked from NousResearch/hermes-agent

    The agent that grows with you

    Python 2