Six months. One author. 948,825 lines of code. AI Engineer & CAIDO. I ship production agentic systems in regulated fintech — and I still write the code.
CAIDO at Altery · 20 yrs TradFi + Crypto rails · Bauman MSTU (First Class) · London · UK Global Talent Visa
Most of my work lives in private repos. Here's the verifiable telemetry.
| Repo | Code LOC | Commits | Last 90d | Stack |
|---|---|---|---|---|
| cosmic-swift (iOS) | 353,212 | 675 | 587 | SwiftUI · Swift · 1,004 files |
| cosmic-copilot (web) | 323,249 | 857 | 427 | Next.js · TS · Supabase · Stripe · OpenAI + Anthropic |
| aeo-platform | 134,178 | 109 | 83 | Next.js · Supabase · OpenAI + Anthropic |
| prime-legacy-homes | 30,002 | 76 | 19 | Next.js · Tailwind · next-intl |
| qaccountants-platform | 23,645 | 1,176 | 67 | NestJS monorepo · Prisma |
| cosmic-landing | 40,828 | 122 | 48 | Next.js · next-intl |
| aeo-landing | 24,040 | 11 | 11 | Next.js · Tailwind |
| qaccountants | 17,808 | 101 | 74 | Next.js · Supabase · PostHog |
| altery-requirements-hub | 1,863 | 109 | 108 | Markdown / docs hub |
| Total | 948,825 | 3,236 | 1,424 | One author on every repo |
That's ~16 commits/day, solo, while running a full-time C-level job. Methodology: find + wc -l excluding node_modules, .next, Pods, DerivedData, build, dist, lockfiles. Read access on request for recruiters and serious inbound.
🔍 AEO Platform — AI-search-visibility engine. Queries Google AI Overviews, ChatGPT, Perplexity, and Claude. Detects hallucinations, citation drift, and source-attribution gaps. Paying users. 134K LOC.
✨ Cosmic Copilot — Chart-accurate AI astrology. VSOP87 ephemeris at ±0.1° precision. 14 languages (~201K words via parallel agents). Native iOS + web, same product, ~676K LOC combined.
🌐 YM8 — Personal site + AI agents showcase.
🏠 Prime Legacy Homes — London residential real-estate platform.
vasilyu1983/AI-Agents-public — public slice of a 192-skill production library across 14 domains with 6-router orchestration. Skills drop into .claude/skills/ or .codex/skills/. Follows the Agent Skills spec.
Notable families:
ai-coding-agents-*(12 skills) — sessions, sandbox, permissions, plugins, observability/evals, terminal UI, remote runtime, command runtime, provider runtime, tools, tasks, release/distribution. Patterns for building durable coding agents.agents-swarm-orchestration— dependency-aware parallel subagents with verifier passes. Used in live operation: 493 sub-agent dispatches, 131 parallel overlaps in a single month, 73% goal-achievement.foundations-game-theory+foundations-decision-theory— agent design primitives. Escalation as Stackelberg games, verifier-builder pairs as repeated games, tool selection as bounded-rational decision under uncertainty.
CAIDO at Altery (UK EMI / FinTech). Built and operate 30 production agents spanning compliance, ops, customer workflows, engineering, and data.
🦅 Hawkeye — AML/CTF horizon scanning, UK + EU-CY jurisdictions 📋 RegLens — Regulatory policy review with gap tables + audit-ready remediation 🔍 KYB Translator — Multi-language KYB doc translation with UBO extraction 🎯 KYB Analyst — AML/CFT risk assessments for business onboarding 📈 Metabase Buddy — Plain English → SQL/MBQL, dashboards in minutes 🔒 Policy Reviewer — Forensic review with verbatim quotes and regulatory citations 🛠️ SmartDeploy · Customer Support · Mobile App Updates · Tech Writer · JD Assistant · DPIA · BPAT · Expansion Scanner · Marketing Researcher + 15 more
Underneath: a 135-repo coordination layer cataloguing 581 API routes across 43 services, 545 documented regulatory gaps across 12 domains, README coverage 123/135, AGENTS.md 17/135.
Agents & AI: Claude Code CLI · Codex CLI · Cursor · v0 · custom MCP servers · OpenAI / Anthropic / Gemini SDKs · RAG · prompt caching · structured outputs · tool use · evals Frontend: Next.js · React · TypeScript · SwiftUI · Tailwind · Radix · React Native · Expo Backend: Node · NestJS · Python · Prisma · C# / .NET (origin stack, still ship) Data & infra: PostgreSQL · Supabase · Firebase · Redis · Vercel · AWS S3 Payments & analytics: Stripe · PostHog · Resend Web3: Solana · Ethereum · Wagmi · Viem · Jupiter DEX Testing: Playwright · Vitest · Jest Fintech depth: KYC/AML/KYB/KYT · Travel Rule · MiCA · DFSA · Visa/UnionPay Principal Memberships · multi-currency card issuing · crypto rails
🎓 MSc Computer Engineering — Bauman Moscow State Technical University (First Class) · BSc Finance (First Class) 💼 20+ years banking tech: Otkritie, Sberbank, VTB, Moscow Exchange, ECOMMPAY — £100M+ IT portfolios, 230+ FTE orgs 🛠️ Production C#/.NET at age 20 — HotSOS hotel CRM, 1,000+ properties worldwide 🌍 London · UK Global Talent Visa · Open to relocation for the right problem
Philosophy: Ship fast, comply hard. Re-architect signal, not noise.
Why the receipt matters
In 2026 everyone claims they "use AI to code." Most can't show the receipt. I run Claude Code and Codex at production intensity daily and the github telemetry above is the externally verifiable artifact: 9 separate repos, 3,236 commits, single author on every commit, ~16/day average.
The point isn't the volume — it's that a senior operator with a C-level day job can still ship as if engineering is the title. That's the "AI-augmented operator" archetype, and I think it's underrated for the next 24 months of agentic AI.
Numbers verifiable on request. Repo read-access available for serious inbound (recruiters, founders, technical partners).