Skip to content
View Universe8888's full-sized avatar
💭
Learning new things
💭
Learning new things

Block or report Universe8888

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Universe8888/README.md
Typing SVG

Lead IT Operations @ DraftKings Inc. · AI Engineer · Python Practitioner

Bridging enterprise IT operations with software engineering, building agentic systems, automation pipelines, and data-driven knowledge tools.

LinkedIn GitHub Email Profile Views


About Me

  • 🏢 Managing global IT operations & procurement at DraftKings Inc. — EMEA fleet, vendor negotiations, ISO 27001 compliance, Freshservice automation
  • 🤖 Building local-first AI agents with reproducible evals — if it can't be benchmarked, it doesn't ship
  • 🌱 Deep in RAG architectures, agentic pipelines, and knowledge graph tooling
  • 💬 Ask me about Python automation, IT asset management, or Obsidian knowledge systems
  • ⚡ Fun fact: I wrote a tool that found contradictions in my own notes — and it was right

✨ Featured Project

wikilens  ·  pip install wikilens

8 shipped agents. Six hand-labeled evals. One Markdown vault.

Turns a folder of notes into a queryable, auditable, self-aware knowledge system.

Stars License Python PyPI

8 agents, with honest per-run benchmark status:

Agent What it finds Measured status
audit Broken wikilinks, orphans, shadowed basenames F1 = 1.00
query Semantic search · dense + BM25 + RRF + rerank Hit@5 = 1.00; Recall@5 0.97-1.00
gap Unanswered questions implied by vault content Cluster recall = 1.00; matcher precision 0.19-0.48
confidence Claims below epistemic threshold (5-level scale) F1 0.66-0.89
contradict Conflicting claim pairs across notes F1 = 0.82
answer Drafts cited stub notes for identified gaps Pass 0.40-0.80; attribution 0.90-1.00
drift Beliefs that shifted over git history Harness + targets; not yet benchmarked
concepts Notes circling an unnamed concept Harness + target; not yet benchmarked

Why it's different:

  • 🏠 Local-first — notes never leave your machine unless you opt into a remote LLM
  • 📊 Evaluated, not vibes — six agents have hand-labeled evals; drift and concepts have harnesses + targets pending published runs
  • 🔀 No vendor lock — swappable embeddings (BAAI/bge), vector stores (LanceDB/Chroma), LLM backends (OpenAI / Claude)
  • 🔍 Hybrid retrieval — dense + BM25 + RRF fusion + cross-encoder reranking; Hit@5 = 1.00 in every mode
  • 🚦 CI-gate readyaudit exits 0/1/2, drop it straight into pre-commit or GitHub Actions
  • 📝 Proof-carrying answers — every drafted note stub carries [^N] citations resolving to vault chunk IDs; attribution rate = 0.90-1.00
pip install wikilens
wikilens ingest ./my-vault
wikilens audit ./my-vault
wikilens query "what did I conclude about X"

🔬 Other Projects

Design-time governance scorecard for MCP servers.

A lightweight scanner for MCP server designs that checks permission boundaries, tool exposure, data handling, and governance readiness before deployment.

Python AI Governance MCP Security Review Policy-as-Code

  • 🛡️ Flags risky tool surfaces before they become production integrations
  • 🧪 Evaluation-backed: 13/13 scan accuracy cases and 5/5 safety properties
  • 📋 Produces reviewable findings for architecture, risk, and implementation teams

Stars Issues

Detect accelerating niche topics before they hit mainstream.

A full-stack trend discovery engine using derivative calculus (velocity & acceleration scoring) on Google Trends data.

Python FastAPI SQLite NumPy SciPy React Vite Recharts

  • 📐 d²V/dt² acceleration as core ranking signal
  • 🌍 8 geo zones — Global, US, UK, EU Big 4, BG
  • 🧪 60+ unit tests across math, DB, API, and ingestion
  • 🎨 Glassmorphism dark-mode dashboard

Stars Issues

AI-powered job scraping with resume matching.

Automated scraper for LinkedIn & Jobs.bg with NLP-driven resume comparison and stealth browser automation.

Python Playwright NLP CSV

  • 🕵️ Stealth browser automation with session persistence
  • 🧠 AI-powered resume-to-job matching scores
  • 📊 Structured CSV export for analysis
  • ⚠️ Built for small-scale educational use

Stars Issues


🛠️ Tech Stack

Languages

Python JavaScript T-SQL Java HTML5 CSS3

Frameworks & Libraries

FastAPI React Vite Node.js NumPy SciPy

AI & ML

OpenAI Anthropic HuggingFace LanceDB

Data & Automation

SQLite Playwright Recharts

Enterprise & Workflow

ServiceNow Jira Microsoft 365 Git


📊 GitHub Stats

GitHub Stats GitHub Streak
Top Languages

🎓 Certifications (2025)

🤖 AI & Generative AI
Certification Issuer Date
AI Agent Developer Specialization Vanderbilt University Jun 2025
Advanced RAG with Vector Databases and Retrievers IBM Oct 2025
Vector Databases for RAG: An Introduction IBM Oct 2025
Build RAG Applications: Get Started IBM Sep 2025
Develop Generative AI Applications: Get Started IBM Sep 2025
Exploratory Data Analysis for Machine Learning IBM Jul 2025
📊 Data Analytics & Visualization
Certification Issuer Date
Advanced Data Visualization with Tableau Tableau Jul 2025
Data Visualization with Tableau Tableau May 2025
Introduction to Tableau Tableau May 2025
Data Ecosystem Tableau May 2025
Business Analysis Process Tableau May 2025
Introduction to Business Analytics Tableau May 2025
Python Data Analytics Meta May 2025
🐍 Python & Systems
Certification Issuer Date
Using Python to Interact with the Operating System Google Jun 2025
Crash Course on Python Google Jun 2025
📦 Supply Chain & Operations
Certification Issuer Date
Leverage Data Science for a More Agile Supply Chain (Specialization) UC Irvine May 2025
Supply Chain Optimization UC Irvine May 2025
Inventory Management UC Irvine May 2025
💻 Software Engineering (SoftUni)
Certification Issuer
Programming Advanced with JavaScript Software University
Programming Fundamentals with JavaScript Software University
Programming Basics with JavaScript Software University
MS SQL (T-SQL) Software University
Microsoft Excel Advanced Software University
Microsoft Excel Fundamentals Software University

🌐 Languages

Language Level
🇧🇬 Bulgarian Native
🇬🇧 English Fluent
🇮🇹 Italian Intermediate
🇩🇪 German Intermediate

Building tools that know what they don't know. 🚀

Plovdiv, Bulgaria · DraftKings Inc.

Pinned Loading

  1. wikilens wikilens Public

    Agentic intelligence layer for Markdown/Obsidian vaults — RAG + evaluated metacognitive agents.

    Python

  2. job-hunt-automator job-hunt-automator Public

    Automated job scraper for LinkedIn & Jobs.bg. Features AI-powered resume matching, stealth browser automation, and CSV export. Built with Playwright.

    Python

  3. pre-viral pre-viral Public

    Detect accelerating niche topics before they hit the mainstream. A full-stack trend discovery engine that uses velocity and acceleration signals to surface "pre-viral" search patterns.

    Python 1

  4. mcp-govcheck mcp-govcheck Public

    Design-time governance scorecard and compliance evidence for MCP servers, audit your AI agent's tool surface against ISO 27001 / SOC 2. Not a gateway, not a firewall.

    Python