I build and operate production agentic infrastructure: MCP servers, agent runtimes, compilers and protocol layers, in Rust, TypeScript, Python and Go. Every figure on this page is measured from the repositories it describes, and each carries the command and the date that produced it.

Agentic AI Systems Engineer & Operator · Production Agentic Infrastructure · AI-Augmented Architecture · Multi-Agent Orchestration

🌐 ishanparihar.com · 🔗 LinkedIn · 📍 Noida, India · ✈️ Remote: worldwide · 📧 Get in touch

Every claim below carries a commit, and every number is re-measured. Start with Hire Me, or read the full catalog and check the code yourself first.

Contents

• 📊 Engineering by the Numbers
• 🎯 Value Proposition — Proof, Not Promises
  • 💼 Hire Me — or verify the code yourself first
• ⚙️ Core Paradigm: How I Operate
• 💎 Flagship Projects
• 🛠️ The Complete Ecosystem
• 🔬 Unified Tech Stack Matrix
• 📂 Complete Project Catalog (56 Ranked Projects + 1 Struck + 6 Web Sites)
• 🧭 The Arc: Space → Mind → Infrastructure
• 📄 Documentation
• ☕ Sponsor this work

📊 Engineering by the Numbers

I design and implement custom cognitive substrates, low-overhead tools, and high-throughput communication routing layers. The operational scope of this portfolio covers:

Metric	Value	Technical Context & Evidence
MCP Servers	21	1,379 registered MCP tools across 21 MCP-exposing repos (portfolio-measurements.json, measured 2026-11-11, re-derived by genome close)
Rust Crates	95	95 Cargo.toml manifests across the 76 published repos (66 federation + 10 catalog) — vendored, generated and documentation trees excluded (portfolio-measurements.json, measured 2026-11-11, re-derived by genome close)
Lines of Code	3.33M	≈3.33M first-party source lines (Rust, TypeScript, Python, Go, Svelte, QML, shell) across the 76 published repos (66 federation + 10 catalog) (portfolio-measurements.json, measured 2026-11-11, re-derived by genome close)
Automated Tests	≈32,000 markers	≈32,000 test markers across 59 of 76 repos; 17 report none because their language has no marker rule here. Markers are NOT passing cases — gate results are recorded separately with their commit (portfolio-measurements.json, measured 2026-11-11, re-derived by genome close)
Context Reduction	40–60%	Token savings verified via custom Token-Oriented Object Notation (TOON) compiler — benchmark claim, not an engine attribute
Production Runtime	<10MB	Stripped static musl Rust binaries with minimal idle RSS, zero dynamic dependencies — benchmark claim, not an engine attribute

All metrics are machine-measured from the repositories — no self-reported numbers. Performance claims (context reduction, runtime footprint) are benchmark-verified on the linked flagships.

🎯 Value Proposition — Proof, Not Promises

I'm an Agentic AI Systems Engineer & Operator — I direct AI-augmented builds of the graph engines, compilers, and protocol layers that make autonomous AI agents functional, resilient, and cost-effective under real production loads. Substrates are chosen per project (Rust, TypeScript, Python, Go, Kotlin) — never as identity. Outcomes over protocols; operator, not coder. Every flagship below points to the proof living in its own repository — playable demos, runnable test suites, and measured benchmarks. Open the repo and verify it yourself.

What that means in measured terms, ranked by the evidence a recruiter can actually open and re-run — the full order is computed by scripts/rank_score.py (rubric: 30% architecture, 25% tests, 20% agent surface, 15% scale, 10% utility, re-derived 2026-11-02), not chosen by hand. All six are public and their gates re-run:

#	Project	Score	The measured proof behind the claim
1	operant	9.40	7,163 passed / 3 failed / 14 ignored across 71 binaries at ca33d4c (2026-10-20 cargo test --workspace) — the highest-test-count runtime in the portfolio
2	switchpost	9.26	1,031 passed / 19 failed across 72 binaries (cargo test --workspace at 2104084, 2026-11-06), 162K LOC across 25 platforms, 389 MCP tools in 44 modules, 3 releases
3	mysterium	8.87	2,110 of 2,112 cases passing across 188 test files (vitest run at HEAD, 2026-10-28), 64-cell developmental matrix
4	openscript	8.70	113 MCP tools, 158 passing / 1 failing, agent-directed brief-to-render pipeline
5	tdg-rust	8.46	626 passed / 0 failed at 44e7f9b (2026-10-20), zero warnings, 50 MCP tools, 10 releases
6	holon	8.10	57K-LOC, 56-command MIT-licensed agent OS — self-healing organization of persistent AI employees, 10 measured sophistication families

Those six carry 12,494 measured test cases between them, each naming its own command and date in docs/FLAGSHIPS.md: 7,163 operant + 1,031 switchpost + 2,110 mysterium + 546 openscript + 626 tdg-rust + 1,018 holon (every figure is passed cases at the named commit; three of operant's 71 binaries, 19 of switchpost's 72 and one of openscript's 20 were red that day, and holon had one failure in 1,018). The portfolio totals 3.46M first-party LOC and ≈31,400 test markers across 75 published repos (see 📊 Engineering by the Numbers). Two private repos score above parts of this table — sourcehound (9.43) and c-suite-agents (8.36, gate red at 76/8/8 of 95) — and a reader cannot re-run either gate today, so both stay in the catalog, locked and capped at B. The claim is not "I can build agents" — it is "I can build agent systems whose behaviour I can prove at a commit, and hand you the proof to re-run."

---

💼 Hire Me — or verify the code yourself first

For founding / forward-deployed engagements, agentic-systems architecture, or fractional CTO work for AI-first teams — 📧 Get in touch. I ship production agentic infrastructure end-to-end via AI-augmented architecture; substrates chosen for the project, not the team. Or audit the code first: start from the portfolio's highest-test-count runtime (Operant, 7,163 Rust tests passing at ca33d4c, 2026-10-20 gate) or the compact fully-verified engine (TDG Rust, 626 tests).

⚙️ Core Paradigm: How I Operate

I treat AI system design as spatial machinery rather than linear text prompting, and I direct its construction via AI-augmented architecture — systems thinking first, substrates chosen per project, the agent runtime writing the implementation under operator supervision. Most agent frameworks today are fragile prompt wrappers. I deliver load-bearing substrates:

• Operator-led, AI-augmented builds: I architect the system, set the test discipline, choose the substrate (Rust/TypeScript/Python/Go/Kotlin), and direct the build. The ≈31,400 test markers across 75 published repos (portfolio-measurements.json, measured 2026-11-06) are the audit trail of that method — not the result of typing each line by hand. Markers are counted, not claimed as passing: every tier row below states its own measured pass/fail at its own HEAD and date.
• Graph-Native Orchestration: I replace linear shell scripts and standard pipelines with property graphs. Every execution node has typed ports; the engine resolves dependencies, materializes parallel branches via a Tokio level scheduler, and exposes the live execution DAG to LLM operators over MCP.
• Token-Efficient Serialization: I designed and compiled the TOON (Token-Oriented Object Notation) standard. Eliminating structural syntax redundancy cuts context-window footprint 40–60% versus JSON, with deterministic validation preserved.
• Minimal Memory footprint: I compile zero-dependency static musl Rust binaries; the intelligence-gathering pipeline curates 411 sources across 47 countries inside a ~7MB static file at <5MB idle RAM.
• Systems Diagnostics: Background in architectural design, spatial diagnostics, and cognitive modeling — I translate systemic risk, organizational constraints, and multi-agent topologies into deterministic software constraints.

💎 Flagship Projects

The ten flagship systems, one line each, in the order scripts/rank_score.py computes — full write-ups with their measured proof live in docs/FLAGSHIPS.md.

Flagship	What it does	Measured proof	Where the proof lives
🔒 📡 sourcehound	Intelligence ingestion across 411 sources in 47 countries — private today (link 404s for visitors; capped at B-tier, re-publication planned)	~7MB static binary, ~5MB idle RSS, 9 custom parsers	the feeds it emits, natively in TOON
🤖 operant	Terminal-native ReAct agent runtime — 60+ tools, hybrid memory, channels	7,163 passed / 3 failed / 14 ignored across 71 binaries at ca33d4c (cargo test --workspace, 2026-10-20)	./scripts/install.sh + operant setup && operant chat — runnable in three commands
🌐 switchpost	Multi-platform content engine across 25 platforms	162K LOC, 389 MCP tools across 44 tool modules, 3 releases, 1,031 tests across 72 binaries at 2104084	the axum REST surface + rmcp MCP server on one binary
🧠 mysterium	Education-system replacement — 64-cell developmental matrix, 9-command CLI	2,110 of 2,112 cases passing (vitest run at HEAD, 2026-10-28), 139K LOC	the 9-command CLI (setup, status, new-game, diagnostic, …) driven from the shell
🎬 openscript	Agent-directed video pipeline — brief → script → TTS → render	113 MCP tools, 158 passing / 1 failing (cargo test at HEAD, 2026-10-28)	docs/ renders produced end-to-end through MCP tool calls
🧠 tdg-rust	Teleological memory graph — every memory a (Content, Embedding, Telos) node	626 tests / 0 failed at 44e7f9b (cargo test, 2026-10-20), 50 MCP tools	the repo's "Testing" section: cargo test = 626 passing (449 lib + 68 MCP e2e + 44 plugin + integration + property suites)
💼 linkedin-lyr	Professional profile extraction + Voyager profile-edit surface	55K LOC, 25 CLI commands + 25 MCP tools, 2 releases	a 25-command agent-first CLI with TOON output and AXI 6/6 ergonomics
🧠 mindstrata	Deterministic emergent society simulation — every agent a complete mind	155K LOC, 307 passing / 3 failing (cargo test at HEAD, 2026-10-28)	cargo test — history always emerges from locally bounded agents
📊 deckmill	Browserless carousel factory — 38 of 46 slide types renderable	192 tests green at 5003ecb, 36K LOC, 20 MCP tools	docs/previews/ — 6 slides actually rendered by the embedded Blitz engine, no browser, no screenshots
🎵 scorestrata	Contract-first music compiler — brief → mastered WAV	944 tests green at a9cb251 (cargo test --workspace, 2026-11-06), 73K LOC	the repo's "Hear it" section: two rendered demo tracks (LP nu-metal, nu-disco) with spectrogram showcases + ss render to regenerate

---

🔒 📡 sourcehound

Intelligence Gathering System — High-performance Rust engine. ~7MB static binary, ~5MB RSS, 143 MCP tools.

🔒 Private today — its link 404s for visitors and its gate cannot be re-run by a reader; capped at B until it is public again. Kept in the portfolio because re-publication is planned; the figures below are from its last public revision, not live claims.

A lightweight news and data ingestion pipeline scraping 411 distinct sources across 47 countries. It filters, deduplicates, and enriches global national, tech, and financial news, generating optimized intelligence feeds for agent consumption.

• Serialization: Outputs natively in TOON, saving 40% to 60% in downstream context window costs.
• Parsers: 9 custom-engineered stream parsers handling RSS, Atom, HTML tables, OFAC lists, and PDF archives.
• Efficiency: Developed to replace a slow, high-memory Node.js prototype, dropping idle RAM consumption to under 5MB.

🤖 operant

Terminal-native ReAct agent runtime. Rust. Tests: 7,163 passed, 3 failed, 14 ignored across 71 binaries at ca33d4c (2026-10-20, cargo test --workspace; the suite is active, so treat the figure as that day's gate, not a current count). MCP client + server, skills, messaging channels.

A production-grade agent runtime written in Rust that runs from your shell: a real think → act → observe loop with a JSON-schema tool registry, memory-context injection every turn, provider fallbacks, and self-healing retries.

• Memory: agentmemory hybrid semantic memory (BM25 + vector + graph), auto-spawned on first use; optional Postgres backend.
• Tools: 60+ JSON-schema tools — fs, git, web, CDP browser, shell, code, http, memory, skills, cron, kanban, notes, checkpoints.
• Channels: Telegram · Discord · Slack · WhatsApp via the gateway; WASM plugins; operant autonomous dev loop.
• Local-first: no telemetry, no account; any OpenAI-compatible endpoint or a local llama.cpp / Ollama model.

🌐 switchpost

Multi-network content ops engine. Rust + SvelteKit. 1,031 passed / 19 failed across 72 binaries (cargo test --workspace at 2104084, 2026-11-06; the 19 are 3 OAuth red-team assertions plus credential-gated live transport tests), 162K LOC, 389 MCP tools across 44 modules, 3 releases.

One binary exposing three interfaces over the same engine: a CLI for shell and agent use, a REST API with a real-time dashboard, and an MCP server for Claude, Cursor and Codex.

• Surface: 389 MCP tools across 44 tool modules, plus a 127-endpoint REST API.
• Platforms: 25 content platforms behind one authenticated interface.
• Gate: build the SvelteKit frontend first (frontend/build/ must exist for the RustEmbed derive), then cargo test --workspace.

🧠 mysterium

Education-system replacement — holonic developmental matrix. TS + Python. 139K LOC, 2,110 of 2,112 cases passing (measured vitest run at HEAD, 2026-10-28).

A structured cognitive curriculum engine: a 64-cell developmental matrix with a holonic growth model, plus a 9-command agent-first CLI (setup, status, new-game, diagnostic, curriculum, session, glossary, profile, setup-profile) — an AXI-style surface agents drive directly from the shell.

• Matrix: 64-cell developmental grid; curriculum, session, and profile state machines over it.
• Surface: mysterium bin globally installable; deterministic, non-interactive commands.
• Skills: ships structured skill libraries (workspace-lint, ui-ux-pro-max, ui-styling) as reusable agent tooling.

🎬 openscript

Agent-directed video editing pipeline. Rust + TypeScript. 113 MCP tools, 546 passing with 1 failing across 20 binaries (measured cargo test --workspace at d9a1924 on a clean export, 2026-11-06; the failure is tools::tests::test_library_search_keyword_query).

A full agent-native media pipeline: take a brief, generate a script, synthesize voices, cut a timeline, and render — driven end-to-end through MCP tools rather than a GUI.

• Pipeline: brief → script → TTS → vision → FFmpeg → Remotion render, orchestrated as MCP tool calls.
• Depth: 12/12 measured sophistication families (state machines, DSLs, protocols, storage, determinism, plugins…) across 5 languages.
• Surface: 113 MCP tools — the widest tool surface in the portfolio after switchpost.

🧠 tdg-rust

Teleological Developmental Graph — Pure Rust cognitive engine. 626 tests (cargo test, 626 passed / 0 failed, measured 2026-10-20 at 44e7f9b; ledger row 40). 50 MCP tools.

A complex implementation of stateful agent memory using a holonic graph structure. Every memory unit is represented as a node: (Content, Embedding, Telos). The teleological decay algorithm ensures stale or irrelevant context decays over time while highly relevant goal nodes remain active.

• Interface: 50 custom MCP tools allowing agents to write, link, prune, and query memories using vector search combined with graph traversal.
• Validation: 626 tests total — zero warnings, zero regressions, verified against the current suite (626 passed / 0 failed at 44e7f9b, 2026-10-20; ledger row 40).
• Canonical: The Rust port supersedes the original Python reference (now private); tdg-rust is the single canonical TDG project.

💼 linkedin-lyr

Professional profile extraction + Voyager profile-edit surface. Python. 55K LOC, 1,400 tests passing, 32 skipped (measured uv run pytest at c629a39, 2026-11-06), 25 MCP tools, 2 releases.

A full LinkedIn integration with a 25-command agent-first CLI and the most complete AXI ergonomics in the portfolio (6/6 axi.md signals). Every interaction is machine-parseable: TOON structured output, a --full escape hatch from content truncation, definitive empty states, size-hinted truncation, pre-computed total_count aggregates, and structured exit codes.

• AXI 6/6: the only portfolio repo implementing every axi.md principle end-to-end.
• Surface: 25 CLI commands and 25 MCP tools; profile fetch, search, and edit flows agents drive deterministically.

🧠 mindstrata

Deterministic emergent human-society simulation — every agent a complete mind. Rust. 155K LOC, 1,258 passing with 3 failing across 12 binaries (measured cargo test --workspace at 6834297, 2026-11-06; the 3 are one simulation-cell crate, the rest of the workspace is green).

Not just a "society sim": a full cognitive-science stack — bodies, nervous systems, appraisal-based emotion, belief systems, social bonds, institutions, culture, and noospheric fields. Three architecture plans (AP1 → AP2 → AP3) deepen it from a medieval settlement into a multi-scale civilization simulation where history always emerges from locally bounded agents.

---

📊 deckmill

Browserless carousel factory. Rust. 192 passed / 0 failed (cargo test --workspace at 5003ecb, 2026-11-06), 36K LOC, 20 MCP tools, 7 releases.

A slide renderer that needs no browser and no screenshots: an embedded Blitz engine draws the slides directly, so the output is a real file a reader can open rather than a claim about one.

• Coverage: 38 of 46 slide types renderable (verified 2026-10-27).
• Proof: docs/previews/ holds 6 slides actually rendered by the embedded engine.

🎵 scorestrata

Contract-first music compiler — from plain-language brief to mastered WAV. Rust. 12 crates, 944 passed / 0 failed (cargo test --workspace at a9cb251, 2026-11-06), 73K LOC.

A deterministic composition compiler that turns a human brief into a full-length WAV: event-sourced composition state, a validator gate with genre/register/layer checks, a production-grade layered synth, and a mastering chain. Same brief → byte-identical audio, every run.

• Pipeline: Brief → plan → compile → validate → render → master.
• Audio: Production-grade layered synth voices (SF2-sampled drums, doubled guitar, organics) with per-layer spectral auditing.
• Determinism: Byte-identical replay from contract + seed; the same 944 tests enforce no-regression rules.

🛠️ The Complete Ecosystem

The operational landscape of these systems, grouped by what they expose. Every row is ranked by the same rubric as the software catalog below, so a tool listed here carries

the score it earned rather than a description alone.

Unified MCP & Agent Infrastructure

System	Stack	Surface	Description
sourcehound	Rust	143 Tools	Lightweight intelligence aggregator scraping 411 sources in 47 countries
switchpost	Rust	389 MCP Tools + REST	Multi-network content ops engine (25 platforms, JWT auth, SSE realtime)
openscript	Rust/TS	113 Tools	AI-directed video editing pipeline (TTS, FFmpeg, Remotion)
artificial-route	Rust	OpenAI + 276 providers	Minimal-RAM LLM gateway — one endpoint over the full catalog, 19 routing strategies, static musl binary
tdg-rust	Rust	50 Tools	Teleological developmental graph memory and knowledge synthesizer
🔒 c-suite-agents	TypeScript	72 Tools	Multi-agent daemon + LifeOS MCP server — private; its gate is red (76/8/8 of 95, bun test at HEAD 2026-10-28) → link 404s, capped at B, not production-grade
holon	Python	56 Commands	Harness-agnostic agentic OS that runs as an organization of persistent AI employees
memory-wire	Rust	54 MCP Tools + REST	Agent memory in one static binary — retain/recall/reflect, bank isolation, FTS5 BM25 + RRF fusion, PII redaction
metatrader-bridge	Python/MQL5	91 Tools	MetaTrader 5 ↔ MCP bridge — regime detection, ATR position sizing, bracket orders, trailing stops
deckmill	Rust	20 Tools	Browserless carousel factory — 38 of 46 slide types renderable (verified 2026-10-27), embedded Blitz renderer
cloakctl	Python	25 Commands	One persistent stealth browser per profile over CLI verbs, plus a registry of reusable agent automations
andrometry	Kotlin/Go	59 Tools	Personal context engine: Kotlin Android collector with a Go MCP server
browsefleet	Node/TS	28 REST Endpoints	Stealth browser fleet: leased CDP contexts, sessions/screenshots/PDFs for agents
automaton	Rust	38 Tools	Graph-native build, plan, and scheduled execution engine
thinking-steroid	TypeScript	13 Tools	Forced epistemic cognitive modes (first-principles, polarity mapping, etc.)
lifeos-bot	Python	Telegram MCP	LifeOS conversational interface — activity/diet/journal/financial logging
lifeos-ops	Rust	31 Tools	Unified CLI + MCP server for Notion-based LifeOS
hermes-prime-bridge	Python	3 Tools	Live plugin importing Prime Agent's persistent kernel into Hermes's stateless runtime

Agent-Native Platform CLIs (-lyr Family) & Obscura

One shared AXI-compliant architecture: every tool speaks TOON/YAML output, structured errors, and contextual help; obscura-core is the shared cookie-vault daemon underneath the browser-scraping family.

Tool	Stack	Surface	Description
linkedin-lyr	Python	25 Tools + 25 Commands	Professional profile extraction and company intelligence scraping
instagram-lyr	Python	47 Tools	Stealth profile reconnaissance and HTTPX-based media scraper
twitter-lyr	Python	45 Commands	Agent-native X/Twitter integration handling DMs, media, and search
facebook-lyr	Python	41 Tools	Facebook / Messenger MCP server via cookie-based scraping
tg-lyr	Python	36 Commands	Fast Telethon-powered Telegram messaging automation client
amazon-lyr	Python	21 Tools	The full Amazon lifecycle as MCP tools — search, cart, orders, shipments, no API keys
etmoney-lyr	Python	10 Tools	ET Money research layer — 67 mutual-fund categories and stock screens as MCP tools
reddit-lyr	Python	56 Tools	Dense Reddit scraping, analysis, and interaction CLI
discord-lyr	Python	16 Commands	Local-first Discord data CLI (SQLite sync, search, analytics)
swiggy-lyr	Python	20 MCP Tools	Unified Swiggy MCP — Food, Instamart and Dineout in one server, one OAuth token
threads-lyr	Python	3 Tools	Threads.net MCP server via anonymous scraping
meme-lyr	TypeScript	6 Commands	AXI-compliant structured image overlay and meme generator
obscura-core	Python	8 REST Endpoints	Stealth browser integration: cookie vault + CDP daemon powering the lyr family

Developer Tooling & Infrastructure — tools the agent-facing work is built with; none exposes an agent surface

Tool	Stack	Surface	Description
android-webcam	Kotlin/QML	1 Command	Android phone as webcam and microphone on the Linux desktop — 720p/1080p, front/back, torch
AI-CAD	TS/Python	23 Endpoints	AI-powered CAD platform — FastAPI floor-plan generation, ezdxf DXF export, real-time collaboration
palimpsest	TS/Python	Library	A training system for one mind — the AI is the trainer, the digestion happens in the reader
project-genome	Python	11 Commands	A project's knowledge genome — owned document roles, mutability classes, and a stdlib CLI (wake/ask/close) — the tool that gates this portfolio's own documentation
linux-telemetry	Python	12 Commands	Production-grade Linux usage telemetry — window activity, browser history, process launches
n8n-compiler	Python	10 Commands	Treat n8n workflows as code
metatrader-docker-deployment	Python/MQL5	5 Commands	Production MetaTrader 4 & 5 Docker deployment with noVNC browser access
pill-bar	QML/JS	Plugin	Omarchy segmented pill bar + static tray. pill-bar is the superset plugin (31 files): segmented-pill-bar (27), pill-suite (29) and static-tray (3) ship the same QML widget set with fewer pieces, and all four are listed individually in the catalog below
omarchy-vpn	QML	Plugin	Omarchy bar plugin for a self-hosted VPS VPN — WireGuard full-tunnel + SSH SOCKS fallback, live stats, IPv6-leak guard

Portfolios & Web User Interfaces (Separate Category — not tiered)

Platform	Tech Stack	Architecture & Operational Details
🔒 ishanparihar-svelte	SvelteKit 5 / Supabase	Core platform with Razorpay integrations, authentication, MDX CMS
🔒 law-of-one-india-website	Next.js 15 / Postgres	Community publishing hub with custom caching, role-based auth, MDX
🔒 design-aesthetics-website	SvelteKit / Tailwind	Architectural portfolio: 49K LOC, GSAP, WebGL shaders, Three.js
🔒 lifeos-website	Next.js 15 / Tailwind	Platform documentation hub and production marketing page
🔒 webdev-portfolio	Next.js / TypeScript	Freelance portfolio scaffold — agent prompt + design system; site source lives in the nested my-portfolio repo
🔒 ishanparihar-cms	TypeScript	Web backend exposing 60+ CMS tools for content management

🔬 Unified Tech Stack Matrix

• Languages: Rust, TypeScript, Python, Kotlin, Go, SQL, Zig, Bash, MQL5
• Protocols: Model Context Protocol (MCP), Agent eXperience Interface (AXI), TOON Serialization
• Infrastructure: Docker, systemd services, PostgreSQL, SQLite, Supabase, LanceDB, n8n automation
• AI/ML Integration: local LLMs (llama.cpp, Qwen), OpenAI, Anthropic, Gemini, Whisper, Whisper Hinglish

📂 Complete Project Catalog (56 Ranked Projects + 1 Struck + 6 Web Sites)

🚀 S-TIER: Flagship Systems — Score ≥ 8.0 — 2 capped by policy (2 private), shown at the score they earned (11 Projects)

• 🔒 sourcehound 9.43 (private): High-speed news ingestion pipeline — 11 sophistication families, 143 MCP tools, 28 releases, 411 sources in 47 countries.
• operant 9.40: Terminal-native ReAct agent runtime — 560K LOC, 7,163 tests passing with 3 failing across 71 binaries at ca33d4c (2026-10-20 gate; the suite is active), 7 workspace crates (re-measured 2026-11-01: Cargo.toml members = [api, os, config, core, cli, plugins, harness], 10 manifests in total — an earlier "19 crates" here counted vendored and nested manifests and is corrected), 12/12 sophistication families (state machines, DSLs, storage, determinism, plugins…).
• switchpost 9.26: 162K LOC across 25 platforms (registry-verified), 10 sophistication families, 389 MCP tools across 44 tool modules, 3 releases, 1,031 passed / 19 failed across 72 binaries at 2104084 (2026-11-06).
• mysterium 8.87: Education-system replacement — 64-cell developmental matrix, holonic curriculum, 2,110 of 2,112 cases passing across 188 test files (measured vitest run at HEAD after svelte-kit sync, 2026-10-28), 10 sophistication families, 9-command AXI-first CLI (5/6 signals).
• openscript 8.70: Agent-directed video editing pipeline — 113 MCP tools, 546 passing with 1 failing across 20 binaries (measured cargo test --workspace at d9a1924 on a clean export, 2026-11-06), 12/12 sophistication families, 5 languages.
• artificial-route 8.52: Minimal-RAM Rust LLM gateway — one OpenAI-compatible endpoint over 276 providers, 19 routing strategies, auto/*, static musl binary. 26 releases, Apache-2.0. Reaches S-tier on ITS OWN measured architecture score (8.6/10), not on a claim.
• tdg-rust 8.46: Teleological developmental graph memory — 626 tests (626 passed / 0 failed at 44e7f9b, 2026-10-20), 50 MCP tools, 11 sophistication families, 10 releases.
• linkedin-lyr 8.44: Professional profile extraction + Voyager profile-edit surface. 1,400 tests passing, 32 skipped (measured uv run pytest at c629a39, 2026-11-06), 25 CLI commands + 25 MCP tools, 2 releases, 9 sophistication families, 6/6 axi.md signals.
• 🔒 c-suite-agents 8.36 (private): Multi-agent daemon + bundled LifeOS MCP server (35 tools, LanceDB/SQLite + Telegram). 76 passing, 8 failing, 8 errors of 95 cases (measured bun test at HEAD, 2026-10-28 — its own gate is red), 10 sophistication families.
• holon 8.10: Harness-agnostic agentic OS that runs as a self-healing, self-evolving organization of persistent AI employees. 57K LOC, MIT, 10 measured sophistication families. Enters at S-tier (8.10) on the same rubric as every other engine here.
• memory-wire 8.08: Agent memory in one static Rust binary — retain/recall/reflect with bank isolation, FTS5 BM25 + RRF fusion, PII redaction on write, over HTTP or as an MCP stdio server. 14 releases.

🧠 A-TIER: Production-Grade Engines — Score 6.5–7.99 — 2 capped by policy (1 experimental, 1 private), shown at the score they earned (18 Projects)

• 🔒 kali-mahabali 7.88 (private): Modular intelligence framework — Codename Chimera. Interchangeable modules (reasoning, memory, tools, models) snapping into a central spine. 63K LOC (Rails engine), CI build, MIT license. Experimental.
• mindstrata 7.82: Deterministic emergent society simulation — 13 crates, 155K LOC, 1,258 passing with 3 failing across 12 binaries (measured cargo test --workspace at 6834297, 2026-11-06), 10 measured sophistication families (state machines, graphs/holonics, DSLs, determinism, plugins…).
• metatrader-bridge 7.67: MetaTrader 5 to MCP bridge — regime detection, ATR position sizing, bracket orders, trailing stops. 91 declared MCP tools across 127 measured REST endpoints, MIT, runs on a 512 MB VPS.
• deckmill 7.66: Browserless carousel factory — 38 of 46 slide types renderable (verified 2026-10-27), embedded Blitz renderer, 20 MCP tools, 192 tests (green at 5003ecb, 2026-11-06), 7 releases, 10 sophistication families.
• instagram-lyr 7.62: Stealth Instagram profile scraper. 368 tests passing (measured pytest at HEAD, 2026-10-30), 47 MCP tools, 5/6 axi signals.
• twitter-lyr 7.51: Agent-native X/Twitter integration. 268 tests passing, 6 deselected (measured pytest at HEAD, 2026-10-30), 45 CLI commands, 0 published releases, 5/6 axi signals.
• holosim-infinite 7.48 (experimental): Math simulator tracking emergent agent interactions — 489K LOC, no measurable test count: the workspace does not compile at HEAD (10 duplicate-definition errors in holonic_realms, cargo build on a clean export, 2026-10-28), 11 sophistication families. Experimental.
• facebook-lyr 7.46: Facebook / Messenger CLI. 295 tests passing (measured pytest at HEAD, 2026-10-30), 41 MCP tools.
• HoloOS 7.30 (spec): Enterprise systems modeling and topological risk — a 14K-YAML + 1.2K-MD knowledge corpus with a holos CLI (187 measured subcommands) and an MCP server. Spec corpus: tracked separately from software, never tiered against the code rubric.
• scorestrata 7.27: WAV music compiler — 12 crates, 944 tests (944 passed / 0 failed at a9cb251, 2026-11-06), 88 MCP tools, 9 sophistication families (state machines, graphs, DSLs, determinism, render/audio…). 73K LOC.
• tg-lyr 7.10: Fast Telethon-powered Telegram automation client. 281 tests passing (measured pytest at HEAD, 2026-10-30), 36 CLI commands, 0 published releases, 5/6 axi signals.
• cloakctl 6.99: One persistent stealth browser per profile over CLI verbs over --json, plus a registry where agents save, compose and compound reusable automations (skills vs workflows). 25 measured CLI commands, MIT, 5 releases.
• amazon-lyr 6.99: The full Amazon lifecycle as MCP tools for agents — search, product detail, reviews, cart, orders, shipments, via browser cookies with no API keys. 21 declared MCP tools, MIT.
• andrometry 6.97: Personal context engine — Kotlin Android collector with a Go REST server. 163 tests passing (measured go test ./... -v at 8d27108, 2026-10-28).
• etmoney-lyr 6.96: ET Money research layer for AI agents — 67 mutual-fund categories, stock screens and NIFTY constituents as 10 MCP tools. Highest AXI score among the newly added repos (4/6), MIT.
• browsefleet 6.85: Stealth browser fleet — CDP pool proxying sessions/screenshots/PDFs to agents. 22 REST endpoints, 86 tests, 4 languages.
• reddit-lyr 6.79: Comprehensive Reddit scraper with 56 MCP tools for agents (browse, post, comment, vote, mod). 86 tests passing (measured pytest at HEAD, 2026-10-30).
• automaton 6.54: Graph-native automation substrate — 17 Rust crates, 38 MCP tools, 8 sophistication families.

📡 B-TIER: Solid Utilities — Score 4.5–6.49 — 4 capped by policy (3 deprecated, 1 experimental), shown at the score they earned (15 Projects + 1 Struck)

• consciousness-fabricator 6.33 (experimental): Audio pipeline compiling dynamic background sounds. 144 passing, 13 failing (measured pytest at HEAD, 2026-10-28) + architecture docs. Experimental.
• 🔒 icode 6.21 (deprecated): Rust command runner parsing agent file edit logs — 11 sophistication families. Archived; its 2,095-test figure predates the archive and cannot be re-measured here.
• jesse 6.08 (deprecated): Deprecated — its own README says so ("no longer actively maintained"). 21 measured CLI commands, MIT. Ranked at its measured score and capped at C by the documented archived/deprecated policy.
• swiggy-lyr 6.02: Unified Swiggy MCP — Food, Instamart and Dineout in one server with one OAuth token. MIT, 3 measured CLI commands.
• discord-lyr 6.02: Local-first Discord data CLI (SQLite sync, search, analytics). 41 tests passing (measured pytest at HEAD, 2026-10-30), 16 CLI commands, 0 published releases, 2 CI workflows.
• thinking-steroid 5.90: Forced cognitive mode engine implementing 13 analysis frameworks. 5 test suites (compiled-JS asserts, measured npm test at HEAD, 2026-10-28 — not a per-case runner, so no case count exists), 13 MCP tools, 5 sophistication families.
• android-webcam 5.80 (infra): Android phone as webcam and microphone on the Linux desktop — 720p/1080p, front/back, torch, mic. Kotlin + QML + Python, MIT. Real desktop tooling (infra category).
• 🔒 lifeos-bot 5.63: LifeOS conversational interface — Telegram bot + agent CLI (simulate/direct/debug). 12K LOC, 33 tests, v0.1.0.
• AI-CAD 5.62: AI-powered CAD platform — FastAPI floor-plan generation, ezdxf DXF export, real-time collaboration. TypeScript + Python + Shell, MIT, 23 measured endpoints.
• ishan-parihar 5.40 (profile): This profile repo — the README, its measurement scripts, and the register that measures every other row. Scored and shown for completeness, tagged profile so no reader mistakes configuration for a software project.
• threads-lyr 5.34: Threads.net MCP server. 57 tests passing (measured pytest at HEAD, 2026-10-30).
• 🔒 lifeos-ops 5.04: Unified CLI + MCP server for Notion-based LifeOS. 31 tools, 8 sophistication families — zero automated tests across 18K Rust LOC.
• CineSync 5.00 (deprecated): Character-asset automation pipeline for lipsync animation. 13.7K LOC, 2 CI workflows. Deprecated — not GitHub-archived (gh api returns archived=false, measured 2026-11-03): the work stopped, the repo is still open.
• palimpsest 4.98: A training system for one mind — the AI is the trainer, the digestion happens in the reader. TypeScript + Python, built on an upstream harness.
• meme-lyr 4.88: TypeScript CLI meme generator. 25 tests, tag-only CI + release pipeline, v2.0.0.
• project-genome 4.54 (tool): A project's knowledge genome — owned document roles, mutability classes, and a stdlib CLI (wake/ask/close) that checks them. 11 measured commands, MIT. The same tool that gates this portfolio's own documentation.

🧰 C-TIER: Operational & Capped — Score 3.25–4.49 (5 Projects)

• obscura-core 4.48: Shared cookie-vault + CDP daemon powering the lyr family — highest cross-repo in-degree (4 dependents).
• workout-factory 4.02: Trainer voice/audio-cache factory. Deprecated — not GitHub-archived (gh api returns archived=false, measured 2026-11-03).
• linux-telemetry 4.02: Production-grade Linux usage telemetry — window activity, browser history, KDE app directories, process launches, desktop-file executions, minimal system impact. Python, MIT, 12 measured commands.
• n8n-compiler 4.00 (infra): Treat n8n workflows as code. 10 measured CLI commands, CI-gated. Infra category — real tooling, no agent surface.
• hermes-prime-bridge 3.83: Live plugin bridging a persistent kernel into Hermes's stateless runtime. 3 MCP tools, 14 contract tests, v0.2.0.

📉 D-TIER: Minimal / Stalled — Score < 3.25 (7 Projects)

• metatrader-docker-deployment 2.79 (infra): Production MetaTrader 4 & 5 Docker deployment with noVNC browser access. Infra category, 5 measured commands, MIT.
• 🔒 lifeos-saas 2.64: Agent hosting stack (NullClaw, Honcho, Postgres). 760 LOC, 0 tests.
• pill-bar 2.05 (tool): Omarchy segmented pill bar — pills only behind widget clusters, tray always expanded, no chevron. QML + JavaScript, MIT.
• segmented-pill-bar 1.98 (tool): Omarchy segmented pill bar — transparent window, pills behind clusters only. QML + JavaScript.
• pill-suite 1.98 (tool): Omarchy pill bar + static tray bundle. QML + JavaScript.
• omarchy-vpn 1.32 (tool): Omarchy bar plugin for a self-hosted VPS VPN — WireGuard full-tunnel + SSH SOCKS fallback, live stats, IPv6-leak guard. QML, MIT.
• static-tray 0.98 (tool): Omarchy static tray — always expanded, no chevron, for segmented pill bars. 48 tracked lines, QML + JavaScript.

## 🧭 The Arc: Space → Mind → Infrastructure

Three disciplines shaped this body of work, and each is visible in how the systems are built — infrastructure for AI agents to operate at production scale: deterministic where it matters, verifiable everywhere, proven by the artifacts committed beside the code.

• Architecture taught me to think in space and load-bearing structure. A program is a container with designed constraints, not a linear recipe — so every repo here is engineered like a building: for what it must carry under real load.
• Psychology taught me to study minds. That's why the portfolio runs deep into cognitive systems — teleological memory graphs (tdg-rust), deterministic society simulations (mindstrata), developmental matrices (mysterium). Building for agents means building for cognition.
• Engineering fuses both into infrastructure: compilers, runtimes, and protocol layers that let autonomous agents operate at production scale (operant, automaton, sourcehound, switchpost).

📄 Documentation

Every doc in docs/ is listed with the question it answers in docs/PAGE-INDEX.md; the ranking engine that produced the tiers above is RANKING-RUBRIC.md and the baseline every repo here is held to is TEMPLATE.md.

Available for remote contract, full-time, and part-time roles worldwide.
📧 Get in touch — let's build the infrastructure of tomorrow.

☕ Sponsor this work

Every sponsorship funds the open ecosystem behind this profile — new releases, test suites, CI, and infrastructure across all 56 ranked projects and 6 web sites: