Skip to content
View spqr-86's full-sized avatar
2 followers · 2 following

Achievements

Achievement: Pull Sharkx2Achievement: YOLO

Achievements

Achievement: Pull Sharkx2Achievement: YOLO

Block or report spqr-86

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
spqr-86/README.md

Hi, I'm Petr 👋

AI Engineer | LLM Systems · RAG · Multi-Agent | Document Intelligence

🚀 About Me

🔭 Building RAG pipelines and multi-agent systems for complex technical documents

🎯 Focus: production eval-driven development — 200+ test cases, LLM-as-judge, Ragas metrics per release

🏗️ Architecture I work with: Router → RAG Agent (ReAct, LangGraph) → Verifier

📫 Reach me: petr.baldaev.ds@gmail.com | Telegram: @PetrBaldaev

💼 Open to AI Engineer roles — Remote / Moscow / International

🛠️ Tech Stack

LLM Systems

Python LangGraph LangChain OpenAI Hugging Face

Vector & Search

Qdrant ChromaDB BM25

Backend & Cloud

FastAPI Docker Google Cloud PostgreSQL

🔥 Featured Projects

Regulatory Compliance Q&A

Production RAG over Russian regulatory documents (GOST, SNiP, Labour Code). Result: 7.7/10 correctness · 93.6% faithfulness · 100% out-of-scope abstain · $0.01/query · 9.5s mean latency

Regulatory MAS Agent

Multi-agent system (LangGraph + MCP) — Coordinator, RegulationsAgent, WebAgent, CriticAgent with self-revision loop.

Financial Budget Assistant

Multi-agent expense tracking via Telegram. Google ADK · MCP · Google Sheets API · Cloud Run CI/CD.

📊 Numbers that matter

Project Key Metric Impact
Regulatory RAG 93.6% faithfulness, 100% abstain on OOS 12× faster search
Water Treatment Analyzer 74% gap-analysis accuracy (domain validated) −40% manual review
Regulatory MAS Self-revision loop · 4 specialized agents MCP tool integration

Pinned Loading

  1. regulatory-rag regulatory-rag Public

    Production RAG for Russian regulatory docs (GOST, SNiP, Labour Code) — 7.7/10 correctness, 0.936 faithfulness, 100% out-of-scope abstain, $0.01/query

    Python

  2. Financial-Advisor-Multi-Agent-Budget-Assistant-v2 Financial-Advisor-Multi-Agent-Budget-Assistant-v2 Public

    AI-powered family budget management through specialized agent collaboration and Model Context Protocol integration

    Python

  3. gitlab-onboarding-rag gitlab-onboarding-rag Public

    A multilingual (RU/EN) RAG system built on 847 pages of GitLab documentation to provide instant answers for new employees, achieving 89% response accuracy and sub-3-second latency.

    HTML

  4. regulatory-mas-agent regulatory-mas-agent Public

    Multi-agent system (LangGraph + MCP) for Russian regulatory Q&A — Coordinator, RegulationsAgent, WebAgent, CriticAgent with self-revision loop

  5. titanic-booking-ai titanic-booking-ai Public

    A conversational AI agent simulating a 1912 Titanic booking clerk, maintaining historical accuracy (94% factual precision) and character consistency through advanced prompt engineering.

    Python