/ PROJECTS

A bare-metal x86 OS kernel written in Rust. Unix-like shell, block filesystem, text editor, BASIC interpreter, and built-in games. Ported from C — the borrow checker caught data races the original never would.

Statically-typed systems language that compiles to C. Ownership-based memory, pipe operators, async/await, and a module system. Built in Go.

Minimal fast logging library for TypeScript. Zero external dependencies — multiple transports, child loggers, async non-blocking writes, structured JSON.

Secure sandbox execution for JS/TS services inside hardware-isolated microVMs. Supports native Firecracker on Linux and Virtualization.framework on macOS.

Extensible NLP framework in TypeScript for fast, structured text analysis. Tokenization, entity extraction, LLM enrichment, caching, and MCP agent integration.

High-performance system metrics CLI in Rust. Monitors CPU, memory, disk, and network with modular design and multiple output formats.

A Game Boy emulator written in Zig. Cycle-accurate CPU emulation, PPU scanline rendering, timer synchronization, and MBC1 cartridge support. Built from scratch — no dependencies.

Zero-dependency HTTP client for TypeScript. 4x faster than axios, 3.7KB, plugin system, and a clean chainable API. Battle-tested in production.

Fast, TypeScript-first string schema engine for validation and pattern rules. Zero-dependency validator with a clean, composable API.

Offline-first school management system. Student records, attendance, grading, scheduling, and analytics — works without internet, syncs when connected.

Small, fast JSON-RPC framework over multiplexed TCP streams. Binary framing, ring buffer flow control, TLS, compression. No Protobuf, no codegen, no dependencies.

A systems programming language for deterministic software featuring linear ownership, compile-time borrow checking, and zero-cost abstractions.

A graph neural network-based recommendation system built from scratch in PyTorch. Implements LightGCN, BPR loss, and 9 advanced techniques including knowledge graph augmentation, temporal gating, and self-supervised contrastive learning.