"Every system should operate like a flight control system — rigorously tested, fault-tolerant, and mission-critical by design."
Building secure, autonomous, and intelligent systems at the intersection of systems programming and AI.
struct Developer {
name: &'static str,
location: &'static str,
roles: Vec<&'static str>,
organizations: Vec<&'static str>,
skills: Skills,
currently_learning: Vec<&'static str>,
currently_hacking_on: &'static str,
}
struct Skills {
languages: Vec<&'static str>,
tools: Vec<&'static str>,
}
impl Developer {
fn eshan_roy() -> Self {
Self {
name: "Eshan Roy",
location: "India 🇮🇳",
roles: vec![
"Open Source Developer",
"Systems Programmer",
"AI Agent Builder",
"Linux Enthusiast",
],
organizations: vec![
"Snigdha OS",
"Tonmoy Infrastructure",
"TIVerse",
],
skills: Skills {
languages: vec!["Python", "Rust", "Go", "Shell", "TypeScript"],
tools: vec!["Linux", "Docker", "Git", "CI/CD"],
},
currently_learning: vec![
"Python",
"Go",
"Rust",
],
currently_hacking_on: "Snigdha OS",
}
}
fn fun_fact(&self) -> &str {
"Be silent when it is time to hear! 🎧"
}
}Most of the work below lives under a small constellation of orgs rather than one personal account:
| Org | Focus | Links |
|---|---|---|
| 🐧 Snigdha OS | The Arch Linux distribution and everything around it | GitLab · GitHub · Dev.to |
| 🌌 TIVerse | AI/ML tooling & dev-infrastructure crates — eclipsera, multigit, dep-insight |
GitHub |
| 🔒 Vision OSS (TIVisionOSS) | Security-focused Rust crates — errcraft and friends |
GitLab |
| 🏗️ Tonmoy Infrastructure | Parent organization backing the security & systems crates | — |
From a full Linux distribution to a growing suite of AI agent tooling — here's what's actually shipped, with real package/download stats pulled live via badges.
🤖 AI & Autonomous Systems
⚡ M31A — Terminal-Native Autonomous Coding Agent
The newest and most ambitious project — a terminal-based AI coding agent written in Go that owns the entire workflow instead of just suggesting code. Every run walks a six-phase pipeline and ends in a verified git commit:
Initialize → Discuss → Plan → Execute → Verify → Ship
| Capability | Detail |
|---|---|
| 🧠 Multi-LLM routing | OpenRouter + Zen providers with parallel health-check fallback and cost-aware model arbitrage |
| 🛠️ Sandboxed toolset | 5 tools (Bash, FileRead, FileWrite, Glob, Grep) behind permission gating, with SSRF & path-traversal guards |
| 📒 Cross-session ledger | Markdown-based learning record — tracks which model performs best per language/framework over time |
| 💭 AutoDream | Automatic context-window consolidation at 60% usage; system prompts are never compressed |
| ↩️ Rollback chain | git-bisect-backed soft/hard/safe resets with automatic backup branches |
| 🖥️ TUI | 29 screens built with Bubble Tea (Elm architecture) — fuzzy model picker, live streaming output |
| 🔐 Key storage | OS-native keychain (D-Bus Secret Service · macOS Keychain · Windows Credential Manager) |
# macOS (Homebrew)
brew install eshanized/tap/m31a
# Linux / macOS (curl)
curl -fsSL https://raw.githubusercontent.com/eshanized/M31A/main/install.sh | bashBeyond M31A, the AI work spans event-driven infrastructure, model fine-tuning, and editor tooling:
| Project | Description | Stats |
|---|---|---|
| neurobus | "Don't send events. Send understanding." A neuro-semantic event bus for cognitive AI systems — transformer-based semantic routing, 4-scope context engine, time-travel event replay, vector memory via Qdrant/LanceDB, and automatic LLM reasoning hooks |   |
| slmgen | Automated fine-tuning engine for Small Language Models (Llama 3.2, Phi-4) built on Python + Unsloth, with a live demo |  |
| NeuroEdit | AI-powered code editor — multi-line autocomplete, natural-language code generation, AI-assisted refactoring, semantic codebase search, privacy-first by design |  |
🦀 Rust Ecosystem (Crates.io)
| Project | Description | Downloads |
|---|---|---|
| eshanized-polaris | Rust-native distributed compute & orchestration framework — pluggable schedulers (RoundRobin/LoadAware), automatic retries with exponential backoff, mTLS by default |  |
| eshanized-polaris-cli | Official command-line interface for the Polaris distributed compute framework |  |
| cynapse | Real-time, memory-resident binary integrity verification — Merkle-tree checksums over executable memory segments, <3% CPU overhead, built for anti-tamper & zero-trust runtime monitoring |  |
| dep-insight | Dependency analysis, auditing & visualization for Rust — duplicate-version detection, license/RustSec auditing, CLI + JSON + HTML dashboard output |  |
| dynpatch | Safe live code reloading for Rust — hot-patch functions, services, and configs at runtime with no restart |  |
| dynpatch-core | Runtime engine behind dynpatch — dynamic library loading, ABI validation, and transactional patching |  |
| dynpatch-interface | Shared interface types & traits for the dynpatch hot-patching system |  |
| errcraft | Structured and colorful error handling for Rust |  |
| multigit | Cross-platform, multi-remote Git sync — push/pull/sync to GitHub, GitLab, Bitbucket, Codeberg & Gitea at once, with daemon mode and OS-keyring auth |  |
🐍 Python Ecosystem (PyPI)
| Project | Description | Downloads |
|---|---|---|
| eclipsera | ML framework built from scratch — 68+ algorithms spanning classical ML, clustering, AutoML & explainability, 100% scikit-learn-API compatible, 618 passing tests. Uses fastalloc object pooling under the hood for a 5–15% speed boost |
  |
| vedart | Parallel runtime unifying Python's concurrency ecosystem |  |
| cynapse | Real-time memory integrity monitor for Python (companion to the Rust crate) |  |
| fastalloc | High-performance Python memory pool library for pre-allocated object pooling — the same engine powering eclipsera's perf mode |
 |
🐧 Snigdha OS - Arch Linux Distribution
A lightweight, Arch Linux-based distro built on the Linux Zen Kernel, tailored for penetration testing 🛡️ and ethical hacking 🕵️♂️
Built for minimalism and user-centric design, the ecosystem ships a dedicated penetration-testing edition alongside general use, and includes Blackbox — a custom framework for managing security tooling. Backed by Tonmoy Infrastructure.
| Project | Description |
|---|---|
| SPM - Package Manager | Snigdha OS Package Manager for streamlined package handling |
| Kernel Installer | Kernel management and installation utility |
| Kernel Manager | Advanced kernel management UI |
| System Maintenance | System maintenance and optimization tools |
| Cleaner | System cleanup and optimization utility |
| Updater | System update manager with GUI |
| Updater TUI | Terminal-based system updater |
| Project | Description |
|---|---|
| Welcome Center | First-run welcome and setup wizard |
| Snigdha OS Tour | Interactive system tour for new users |
| Assistant | AI-powered system assistant |
| Raptor | System configuration and management tool |
| Wallpaper Engine | Dynamic wallpaper management system |
| Project | Description |
|---|---|
| i3 Config | i3wm configuration and rice |
| Hyprland Config | Hyprland compositor configuration |
| GNOME Config | GNOME desktop configuration |
| Plasma Config | KDE Plasma configuration |
| Aurora Theme GTK | Custom GTK theme for Snigdha OS |
| SDDM Sugar Candy | Custom SDDM login theme |
| Project | Description |
|---|---|
| NERVE | TUI-based system installer |
| Calamares Aurora | Graphical system installer |
| Archiso GNOME | GNOME edition ISO builder |
| Archiso i3 | i3 edition ISO builder |
| Package Browser | Web-based package search and browser |
| Project | Description |
|---|---|
| Snigdha OS Web | Official website source |
| Documentation | User and developer documentation (Docusaurus, multi-language) |
| Tools DB | Security tools database |
🌐 Web & Tools
| Project | Description | Tech Stack |
|---|---|---|
| eshanized.github.io | Personal portfolio site — Next.js/React, OneUI-inspired design |   |
| eportfolio | A second portfolio build exploring a different visual direction |  |
| resume | Open-source, JSON-driven personal resume — fork it, edit resume.json, deploy free on GitHub Pages |
 |
| i3wm | Custom i3wm configuration and dotfiles |  |
- 🤖 Building an AI Agent That Learns from Its Mistakes: The Ledger System
- ⚡ Building M31A: A Terminal-Native AI Coding Agent That Ships, Not Just Suggests
- 🧪 I Built an Automated SLM Fine-Tuning Engine with Python and Unsloth
- 🔀 MultiGit: One Repository, Infinite Destinations
- 🦀 Unlock Clearer Rust Dependency Insight with dep-insight
➡️ Read More Articles on Dev.to
🤝 Collaborations · 💡 Open Source Projects · 🎤 Speaking Engagements · 📚 Mentoring · 🐧 Linux & Systems Development
⭐ From eshanized with ❤️