Projects
Things you can run in a browser first, then the code behind them.
Pinned

nl2sql playground
Ask your database questions in English. The model emits a typed query plan, never SQL text - validated against the real schema and the caller's role before any SQL is generated.
question → planner (typed plan, never SQL) → validator (schema, joins, role) → sqlglot → read-only executor
A plan that fails validation never becomes SQL, and a security refusal is final. On PyPI as nl2sql-engine.

Post-training
A collection of small, runnable implementations of LLM post-training and alignment methods, from RL basics to DPO, RLHF, and RLAIF.
environment → agent (Q-learning, DQN, PPO, GRPO) → policy update → tests
Each method is small enough to read in one sitting, and each has its own tests.

Decision Arena
Decision Arena: TypeSafe's Jev vs open-source Laya playing highway-env, Snake and Blackjack with zero training, plus benchmarks and a Claude Code watchdog
game state → decision model (Jev or Laya) → probability per allowed action → environment step
The model can only answer with an option on the list. Includes agent-watchdog, which scores each Claude Code tool call before it runs.

RAG playground
A local-first bench for learning and demonstrating RAG by experiment
PDF → parse → clean → chunk → index → retrieve (dense, keyword or hybrid) → rerank → cited answer
Every stage is swappable, and evaluation says why each miss missed.
- logscribe (opens in a new tab)
AI-powered log analysis for Python logging: batch, scrub PII, and route logs to an LLM for insights.
Python, Aug 2026
- medalflow (opens in a new tab)
dbt, but in Python classes. Declare medallion (Bronze/Silver/Gold) models as Python classes; MedalFlow extracts dependencies from your SQL and compiles them into a staged execution plan.
Python, Aug 2026
Other repositories
- tagalong (opens in a new tab)
A crew of AI companions that tag along on your screen: they see what you see, answer by text or voice, remember, talk to each other, and act through MCP tools you approve. Never mouse or keyboard control.
TypeScript, Nov 2026
- neural-arcade (opens in a new tab)
Small browser games where tiny neural networks learn in front of you. Each teaches one machine learning idea.
JavaScript, Nov 2026
- home_server (opens in a new tab)
Turn old laptops into Ubuntu home servers from one USB stick: Docker, Tailscale, Cockpit, firewall, battery limit.
PowerShell, Nov 2026
- ai-experiments (opens in a new tab)
Measured experiments on decision models. Can an open-weight model re-rank search results? Does a token budget explain a benchmark failure? What does one typed decision cost across model families? Every request and response is recorded, and negative results are published as they came out.
Python, Nov 2026
- llmwarden (opens in a new tab)
Per-stage model settings, secrets by reference, guarded endpoints and key hygiene for AI applications. Proposal.
Python, Oct 2026
- doc_generator (opens in a new tab)
A tool to automatically add docstrings to your Python code using Large Language Models (LLMs), ensuring your code remains unchanged except for the addition of docstrings.
Python, Jun 2026
- mini-gpt (opens in a new tab)
A compact, readable GPT-style decoder-only Transformer implemented in pure PyTorch. The goal is to expose the essential architectural pieces with minimal scaffolding so you can train and tinker quickly.
Python, Nov 2025
- search_algorithm (opens in a new tab)
Implements the BM25 search ranking algorithm two ways (via the rank_bm25 library and from scratch) and compares the resulting scores.
Python, Sep 2024
Earlier work, before 2024
- microservice_demo (opens in a new tab)
This is a simple microservice demonstration using spring boot.
Java, Aug 2021
- spring_boot_multi_module_framework (opens in a new tab)
This project aims to help the developer to quickly bootstrap multi-module project.
Java, Jul 2021
- chess_engine_using_python (opens in a new tab)
Chess Engine Implementation using Minmax, Alpha-Beta Pruning, and Quiescence Search Algorithm.
Python, May 2021
- rain_in_austrailia_kaggle (opens in a new tab)
Kaggle Kernel for Rain In Australia dataset. It predicts whether there will be Rain tomorrow in Australia or not with 85% Accuracy.
Jupyter Notebook, Apr 2021
- vehcile_tracking_system (opens in a new tab)
Solidity smart contracts ("Floorchain 2.0") modeling auto-dealer floorplan financing — lender agreements and vehicle inventory/state tracking.
May 2019