DataKollekt AB — Data platforms and AI agents that run in production
DataKollekt AB is the Stockholm engineering consultancy of Pooya Shahryari — 7+ years building data platforms for Swedish enterprises, now building AI-agent systems that stay in production. Omvandlar data till resultat.
Services
Data engineering
ETL/ELT pipelines, data warehousing (Data Vault 2.0, dimensional), and migrations off legacy stacks. Tools: SQL, Python, Pentaho, SSIS, Azure Data Factory, Snowflake, PostgreSQL, Oracle, Elasticsearch, Power BI / Qlik, Apache Airflow, Apache Spark, Databricks.
AI-agent & LLM systems
Production agents on the Claude Agent SDK and MCP: autonomous services, custom MCP servers, LLM pipelines with test suites, CI gates, sandboxing and secrets management — engineering discipline applied to AI.
Automation & platform operations
Self-healing services, infrastructure-as-code, monitoring, disaster-recovery runbooks, privilege-separated security models, and Linux. Systems that keep running when nobody is watching.
Selected work (internal R&D)
- Quant — advisory-only AI trading cockpit; FastAPI + SSE backend, React 19 frontend; the AI has zero execution authority; 150+ test files with CI-enforced architecture.
- Agent platform infrastructure — infrastructure-as-code for a multi-agent stack running 24/7 on self-hosted hardware, with privilege separation, keychain-only secrets, and a commit-gate pipeline.
- Content engine — a fleet of hand-built MCP servers and inference/media services; 400+ test files.
- Stealth SaaS — a multi-tenant AI teammate for software teams (in development).
- Open source: voice-inbox-bot; spotify-genre-sync.
Founder — Pooya Shahryari
Data engineer turned AI-agent engineer. 7+ years building data platforms for Swedish enterprises; since 2024 investing in a production multi-agent AI platform running 24/7 on self-hosted infrastructure. DataKollekt is deliberately a one-person consultancy — you work directly with the engineer who built everything.
Experience
- DataKollekt AB — Founder & consultant (2021–present), Stockholm
- Dun & Bradstreet (Bisnode) — ETL developer: consolidating ~400 ETL jobs from a dozen legacy technologies into Pentaho (Java, Perl, Bash, PostgreSQL, Oracle, Elasticsearch)
- XLENT — Data engineer (2019–2021): Data Vault 2.0 warehouse on Azure for Svenska kyrkan; SQL Server DW at Wasa Kredit
- Avaus — Data engineer (2019): customer analytics for Axfood (Python, Snowflake, Teradata)
- Academic Work — Java developer (2018–2019)
- DIW — Founder (2014–2019): e-commerce
- Huddinge kommun — Team lead / acting unit chief (2012–2018)
Certifications: Certified SAFe 6 Agilist; Professional Scrum Master (scrum.org). Working languages: Swedish, English, Persian.
Contact
Email: pooya@datakollekt.se · LinkedIn · GitHub · DataKollekt AB, Stockholm, Sweden.