Как мы строили MVP data lineage системы в ЮMoney

Привет, меня зовут Юля, я дата-инженер в департаменте аналитических решений ЮMoney. Мы разрабатываем и поддерживаем ETL(ELT)-процессы загрузки данных для BI-продуктов компании: отчётов, дашбордов, витрин, API и других. В этой статье поделюсь, как мы разрабатывали MVP для построения data lineage системы и к каким выводам пришли, — так вы сможете адаптировать подход под свои процессы. Проблема, которая привела нас к разработке MVP, заключается в том, что в начале работы над большинством задач требуется подготовительная аналитика: восстановить путь происхождения данных, чтобы понимать, какие ETL(ELT)-процессы будут затронуты и на что повлияют изменения. Это увеличивает время выполнения задачи. Но, что ещё существеннее, процесс достаточно рутинный: каждый раз необходимо проводить статический анализ кода и изучать документацию (при её наличии). Мы решили автоматизировать эту работу, чтобы иметь под рукой инструмент для построения пути данных по точкам их обработки и применения, — то есть создать data lineage систему.

https://habr.com/ru/companies/yoomoney/articles/1027068/

#data_lineage #etl #elt #dwh #neo4j #kafka #cypher #data_governance #mvp #метаданные

https://neo4j.com/

https://proxify.io/

#neo4j

Impact.com is hiring Senior Data Scientist, Product Data

🔧 #python #cicd #elasticsearch #gcp #neo4j #seniorengineer
🌎 Cape Town, South Africa
⏰ Full-time
🏢 Impact.com

Job details https://jobsfordevelopers.com/jobs/senior-data-scientist-product-data-at-impact-com-feb-13-2026-bfb995?utm_source=mastodon.world&utm_medium=social&utm_campaign=posting
#jobalert #jobsearch #hiring

The other fun project of the weekend was another kind of beast:

I wanted to be able to listen system output from specific RPG programs running on our AS400

The idea was the following:

could I generate a graph from dependencies used when X or Y RPG routine run?

I wanted to be able to map physical programs from libraries and other dependencies.

so I went the Java path since its the main language we use to interface with AS400 routines.

Then I used Neo4J to graph the output of the Java ibmi listening server.

I’m probably going to write a long form about this since it was really interesting to vibe code.

At some point I had to query the RPG routine themselves to be able to find key dependencies my main RPG routines were using.

so I basically created an API endpoint that I can query with specific program names to then find their dependencies and again graph the results to Neo4J

I still have to fine-tune the codebase before I publish this publicly but I’m pretty happy about the results

🔗 https://rmendes.net/notes/2026/04/19/60bd6

Latent.Space (@latentspacepod)

Neo4j의 에밀레 프리프가 현대 AI에서 그래프와 컨텍스트 그래프의 역할을 다루는 강연을 다시 소개한다. 문서를 지식으로 바꾸는 그래프 기반 접근이 다시 주목받고 있으며, 2년 만의 업데이트 내용이 담긴 영상이다.

https://x.com/latentspacepod/status/2045648542166106388

#neo4j #graphai #contextgraphs #knowledgegraph #ai

In my preparation of my talk @jcon about #JSpecify, I released version 4.0.0 of #Neo4j migrations, fully embracing JSpecify and #NullAway. Null-safety on the Java module path, as it should be. See you next week in Cologne.

https://github.com/michael-simons/neo4j-migrations/releases/tag/4.0.0

Reddit is hiring Staff Machine Learning Systems Engineer

🔧 #python #gcp #kubernetes #neo4j #terraform #machinelearning
🌎 Remote; United States
⏰ Full-time
🏢 Reddit

Job details https://jobsfordevelopers.com/jobs/staff-machine-learning-systems-engineer-at-redditinc-com-mar-18-2026-61f28e?utm_source=mastodon.world&utm_medium=social&utm_campaign=posting
#jobalert #jobsearch #hiring

🕸️ AI Diabetes Coach—Knowledge Graph

Every AI conversation starts from zero — no memory of your history, your patterns, what spikes you, what doesn't.

A knowledge graph fixes that.

Part 4 of my AI Diabetes Coach series: 🕸️ Knowledge Graph

https://blog.warrenweb.net/ai-coach-part-4

#KnowledgeGraph #Neo4j #ContextGraph #PKM #AI #diabetesmanagement

🚀 Part 2: AI Diabetes Coach System Architecture

From 82% → 98% glucose control using:
- Obsidian (PKM)
- Neo4j (knowledge graph)
- Claude AI (coaching)
- Python (automation)

5 components working together. 9 screenshots. Complete technical breakdown. $20/month.

Real implementation, not theory.

https://blog.warrenweb.net/ai-coach-part-2

#AI #HealthTech #PKM #Obsidian #Neo4j #Automation

Parts 3-5 coming: Prompts, Queries, Build Your Own.

Graph DB vs relational DB for AI apps… is it hype or actually useful?

We’re refactoring a real app live and testing it ourselves. Curious where graphs actually shine. Jump in and help us break it 🎭
https://youtube.com/live/KUcZL2yHjGU
#Neo4j #GraphDatabases