Hands-on senior software engineer and software architect with extensive experience building and operating production systems—from backend, web, and mobile development to AWS architecture, CI/CD, observability, reliability, and performance optimization.

I work primarily with Java/Spring, Node.js, Python, React, React Native, and Angular. I enjoy staying close to both the customer and the production experience: understanding real customer needs, translating them into valuable products, and owning the engineering journey through implementation, deployment, and production operations.

I particularly enjoy investigating complex production incidents, finding their root causes, and turning those findings into simpler, more reliable, observable, and maintainable systems.

• Event-Driven Architectures (EDA)
• AWS IoT platforms and telemetry pipelines
• Serverless performance & cost optimization
• Distributed system reliability and observability

• Distributed systems design
• System behavior under load and concurrency
• Observability and production diagnostics
• Performance engineering and reproducible load testing
• Failure-mode analysis in cloud systems

My public projects span two areas: practical tooling for testing, investigating, and operating distributed systems in AWS environments, and explorations in multi-agent AI orchestration and generative applications.

AI-powered interactive animated story generator built on multi-agent orchestration and deterministic validation.
Uses a pipeline of four specialized Claude agents on Amazon Bedrock to generate complete branching episodes — story script, character choreography, SVG obstacle art, and frame-by-frame animation — from a single natural language prompt.

AWS-native incident investigation workflow built around deterministic workers and bounded AI.
Uses Step Functions to orchestrate evidence collection across metrics, logs, and traces, while GenAI serves as an advisory layer to compare competing hypotheses, interpret cross-source evidence, and surface missing evidence.

Generic AWS workload generator built on ECS Fargate with pluggable scenarios (IoT, SQS).
Designed for controlled load generation and analysis of system behavior under stress.

Repeatable k6 load testing framework running on ECS Fargate, generating traffic from a consistent cloud environment instead of developer laptops.

• Load generation and reproducible performance testing environments
• System behavior under high concurrency and burst traffic
• Event-driven system reliability patterns

• When a few noisy devices took down the system: lessons from a production investigation
  A production incident investigation from an IoT canary OTA rollout that exposed how a small subset of noisy devices, hot database paths, app-triggered follow-up API calls, and retry behavior combined into system-wide contention.

• Turning Noisy AWS IoT Presence (connect/disconnect events) into Reliable Connectivity State
  How to turn noisy IoT connect/disconnect streams into a bounded, reliable connectivity signal using rate limiting, back-pressure, and derived state instead of treating raw presence events as truth.

• How do you find the cost vs performance sweet spot of an AWS Lambda function?
  Exploring Lambda memory tuning impact on cost and latency, validated with AWS Lambda Power Tuning and controlled load tests.

• AWS IoT Services Deprecation: From Managed Pipelines to Modular Cloud Architectures
  What the retirement of IoT Analytics and IoT Events signals about AWS’s direction for modern IoT architectures.

• Practical takeaways for reducing AWS Lambda cost
• Going beyond Power Tuning: validating cost vs latency with controlled load tests (k6 on Fargate)
• How did we reduce AWS cloud costs by ~50%? Practical actions + tools

• Private subnet connectivity: NAT vs Gateway Endpoint vs Interface Endpoint

• Most production incidents aren’t hard to fix — they’re hard to understand fast enough
  A monitoring maturity journey and lessons from improving incident response.

• Sometimes the best architecture improvement is removing an unnecessary path
  Simplifying a production IoT progress-update flow by replacing a heavier event-processing path with a direct IoT Rule → Device Shadow pattern, reducing latency, cost, and operational complexity.

• Load-test your EDA system without permanent infrastructure

• Golden AWS IoT insights for reliability, scale, and cost-efficiency

• AWS IoT Core as a modular event-flow layer — routing, rules, shadow state, and beyond

• Migrating on-prem to AWS is a maturity journey
  Business goals first, cloud as leverage.

• Analysis paralysis vs bias for action

• Amazon AgentCore: production AI agents at scale (AWS Israel event takeaways)

• LinkedIn