Agentic AI · Enterprise Architecture · Data

Kayvon Salari

Senior data and enterprise architect with 25+ years of experience building large platforms for European and global organisations. Now building multi-agent AI systems for real enterprise problems.

25+
Years in enterprise architecture
Sky Germany · Lufthansa · Bayer · HP · Deutsche Telekom · Volkswagen
12
Agentic AI systems built
LangGraph · LangChain · RAG · A2A Protocol · Databricks · Reinforcement Learning
3
Cloud certifications
GCP · Azure · TOGAF 10 · SAFe for Architects
JHU
Certificate Program in Agentic AI
Johns Hopkins University · Completed · July 2026

Multi-Agent Systems

Building now, with new systems and code going out regularly.

Investment Intelligence · Multi Strategy
Aristos Council: Equity Research You Can Audit

An equity research system based on one idea: AI you can audit. A company is compared with peers from about 25,000 listed companies and checked with a set of stock strategies, each asking one simple question: cheap and good, growing at a fair price, steady income, durable business. A separate check asks whether the profits are real. The result is fully deterministic: the maths decides, the AI writes, and a fact checker reads every sentence against the numbers. Every strategy has a ten year backtest record per industry, and a blind three model judge panel gives a second opinion with no vote. Public repo, more than 5,000 tests.

LangGraphMultiple AgentsyfinanceEODHDBacktestingFact Check PassJudge PanelLangSmithGitHub Actions
Financial Services · Platform Engineering
Governed Document to Decision Agent on Databricks

Mortgage paperwork goes in; six figures come out, each tied to the document it came from, with anything that does not add up flagged. The decision never touches the model: a Unity Catalog SQL function applies the lending rules from a policy table, and anything short of a clean pass is sent to a person with a drafted case note. I then tested it four ways. Three attacks failed; a note hidden in a real payslip fooled the AI on the first try. The README is a clear field report on why, and on the code checks that would stop it. Served as an MLflow endpoint and traced end to end.

DatabricksUnity CatalogMLflowModel ServingLlama 4 MaverickAttack TestingDeterministic Rules
Investment Research
AI-Powered Investment Research Platform

Analysts spend hours comparing market data with strategy documents. DualLens ranks companies on both at once — three years of stock performance combined with insights RAG-extracted from strategy PDFs — so it is clear which names are financially strong and ready for AI adoption. One query replaces the manual comparison loop.

RAGLangChainChromaDBYahoo FinanceOpenAI
Investment Intelligence
Autonomous Financial Research Analyst

Per-company investment research takes an analyst 4–6 hours; this LangGraph agent does the gathering in minutes. It pulls real-time market data, news sentiment, and private analyst reports, then gives a Buy/Hold/Sell recommendation with every source cited — easy to check and fast enough to run across a watchlist. The predecessor to Aristos Council, where its single-agent verdict grew into a gated, deterministic decision core.

LangGraphLangChainTavilyRAGOpenAI
Financial Services
Senior Mortgage Underwriting System

Mortgage underwriting takes 3–5 days mainly because four checks — credit, income, assets, collateral — wait one after another. This system runs them as specialist agents under a Supervisor, cutting the cycle to hours while keeping the parts regulators care about: Fair-Lending compliance, PII protection, bias detection, and mandatory human-in-the-loop escalation on every contested file.

LangGraphMulti-AgentRAGChromaDBHITL
Clinical Decision Support
MS Risk Screening Agent

Early multiple-sclerosis signals are spread across EHR records, where no one has time to connect them. This multi-agent system scans records and flags patients showing early risk patterns for neurologist review — with adjustable autonomy, a clear reason for every flag, and PHI governance throughout. It surfaces candidates for clinical judgment; it does not replace it.

Multi-AgentHAIEHR SimulationResponsible AILangChain
Research Intelligence
AGI Research Intelligence Platform

Keeping up with AGI-relevant research on arXiv takes weeks by hand and still gives partial coverage. A Planner→Discovery→Evaluation agent pipeline automates it, scoring every paper against a standard 10-parameter AGI framework — literature review in hours, with 3–5× the coverage of manual survey.

LangGraphMulti-AgentarXiv APIOpenAI
Content Intelligence
LinkedIn Post Generator

A content pipeline with editorial standards built in: a Researcher gathers material, a Writer drafts, a Critic scores against a quality rubric, and a Supervisor repeats the loop until the draft clears the bar. The key part is the loop — generation that does not ship until an adversarial agent approves it.

LangGraphMulti-AgentLangChainOpenAI
Healthcare · Data Governance
Healthcare Intelligence Assistant

Clinicians wait on DBAs for every data question. This Natural Language-to-SQL system lets them ask directly — and keeps it safe by classifying every query before execution: READ runs automatically, WRITE needs human approval, UNSAFE is rejected outright. Full audit trails keep it inside HIPAA and GDPR.

LangGraphNL-to-SQLHITLHIPAAGDPR
Public Sector · Knowledge Management
MucAtlas — Municipal Knowledge Agent

City caseworkers lose time searching across fragmented internal sources of unknown currency. MucAtlas (Munich Innovation Challenge 2026 entry) is a five-agent RAG system that connects those sources, checks whether legal documents are still current, finds contradictions between sources, and returns cited answers in seconds — with full data sovereignty via confidential computing. Architecture and design; competition entry.

Multi-AgentRAGConfidential ComputingMCPGDPR
Logistics & Robotics
Warehouse Robot Navigation

Programming warehouse robot routes by hand does not scale past the first layout change. This reinforcement-learning agent (PPO) learns goal-directed navigation from scratch and generalises to layouts it has never seen — pathfinding as a learned skill rather than a maintained ruleset.

Reinforcement LearningPPOOpenAI GymStable Baselines 3
Sports Analytics · Agent Protocol
Tennis Oracle — Deterministic Match Prediction

Betting markets price the top of the game well and ignore the bottom entirely. Tennis Oracle rates every ATP and WTA player from a million matches — tour, Challenger and ITF — and predicts at 65% where the market sits at 68%, but at 70% on women's ITF events no bookmaker prices at all. Verdicts come from surface-adjusted Glicko-2; the LLM only narrates. Speaks the A2A agent protocol, and a blind test of 197 matches showed the model analyst losing to the ratings by 7.6 points, which is why it narrates.

A2A ProtocolGlicko-2Multi-Source FeedsStreamlitWalk-Forward Backtest

I've spent 25+ years designing and delivering large-scale data platforms and enterprise architecture across some of Europe's largest organisations — Sky Germany, Lufthansa, Bayer, HP, Deutsche Telekom, and Volkswagen, among others.

As Head of Data Architecture at Sky Germany, I built two international teams from scratch, led the migration of on-premise infrastructure to the cloud, drove GDPR compliance across data systems, and delivered unified data products and cataloging capabilities — eliminating 100% dependency on external consultancies.

I co-founded Zeal Hub GmbH, a boutique data and BI consultancy, running enterprise data platform engagements independently for five years.

I completed the Certificate Program in Agentic AI at Johns Hopkins University (Whiting School of Engineering), and now apply that foundation to build multi-agent systems that solve real enterprise problems across finance, healthcare, and public sector.

◈
TOGAF 10 Certified
Enterprise Architecture · Sep 2024
◈
Google Cloud Certified
GCP Professional Data Engineer
◈
Microsoft Azure Certified
Azure Data & Cloud Architecture
◈
SAFe for Architects
Scaled Agile Framework
◈
Certificate Program in Agentic AI
Johns Hopkins University · July 2026

Based in Europe,
looking at new opportunities.