01 / noise
Large-scale event data landing from Kafka, Event Hubs, S3 and half a dozen databases. Nobody trusts it yet, and nobody owns it. I start here, usually with the analysts and product people who need something out of it.
02 / rhythm
A live streaming ETL pipeline on Spark and Databricks: 10,000 events a second at peak, hundreds of terabytes a month, Delta Lake underneath, Terraform and CI/CD keeping it honest. I built it, and I'm who the team asks when something in it breaks.
03 / harmony
An AI agent that works like an analyst: it understands company-specific domain data and runs real analysis workflows - about 1000× faster than a human analyst, at 1/100 of the cost. It wasn't on anyone's roadmap. I scoped it, built it, and pitched it until it was.
04 / performance
Frontend, backend, security and evaluation - from concept to a product used internally and pitched to customers. I wrote the code and ran the conversations with product, sales and security that turned it into an offer we win with.
2022 - today
2025 - 2026
1000× faster · 1/100 the cost
CYMOTIVE Technologies
LLMsLangChainLangGraphRAGMLflowPython
Initiated and independently built a new AI product from concept to proof-of-concept: an AI agent that works as an analyst over company-specific domain data - about 1000× faster than a human analyst at roughly 1/100 of the cost.
2023 - 2025
10k events/s · 100s of TB/month
CYMOTIVE Technologies
DatabricksApache SparkDelta LakeKafkaTerraformSQL
Built a live streaming, end-to-end ETL pipeline that consumes 10,000 events per second at peak - hundreds of terabytes every month - on Spark and Databricks, and led a full system migration from Elasticsearch into Databricks.
2022 - 2023
End-to-end features
CYMOTIVE Technologies
TypeScriptNode.jsVueReactPostgreSQLAWSAzure
Developed backend services, REST APIs and data-driven applications in TypeScript and Node.js, with Vue and React frontends and serverless functions on AWS and Azure.
An SQL interface for managing containers. Jagura behaves like any SQL database, plus a CONTAINER data type: start, stop, restart, pause or kill containers, read their metadata and run commands inside them - all from ordinary SELECT statements.
An AI-based competitive intelligence platform. It tracks competitors, turns their product, pricing and security changes into insights, compares them side by side and ranks everything by impact automatically - with every claim traced back to its source.
One agent skill per engineering book I've read - each distilling the book's rules and methodology into a single SKILL.md, so I can call a book into my agent while developing. Covers AI engineering and RAG, data-intensive systems, refactoring, pragmatic engineering and UX. Works in Cursor, Claude Code and anything else that speaks the open Agent Skills format.
An article I wrote after benchmarking PySpark against Scala: identical performance on the native DataFrame and SQL APIs, but up to 10× faster in Scala once heavy UDFs enter the picture - and the rewrite that finally stopped a streaming job from running its driver out of memory.
About
Data and AI Engineer with a strong backend, data engineering and cloud infrastructure background. I've worked the whole product lifecycle - frontend, REST APIs, large-scale Spark pipelines, cloud infrastructure, AI agents, RAG and context engineering - and I like taking a proof-of-concept all the way to a product people actually use.
ENGLISH EXCELLENT · HEBREW NATIVE