ML LAB

Focusing on efficiently utilizing Agentic AI tools and workflows from 2 different perspectives:
1.	Integrating existing apps/services/APIs as MCP servers to Agentic workflows. Exposing MCP compatible APIs from the apps to be consumed by Agents.
MCP server is build using .net mcp SDK. 
The server tested by Copilot studio, Claude Desktop, GitHub copilot and Cline.
Thus, you can get app data to agent sessions and agents can act, by executing actions on the app! 
You can try it yourself! More in the link bellow or visit todoapp.mllab.nl
2.	Making sure AI Coding agents can contribute efficiently to the project: Considering coding agents as Codex, Claude Code, Cline, Copilot for local development. Undelaying LLM providers and their cost structure: OpenAI, Anthropic, Azure AI Foundry, OpenRouter. Using less capable models to execute on plans created by more capable models. Making GitHub Copilot an efficient and safe code contributor! Make fully automated process with several automated check - compilation, Unit and Integration tests for both Frontend and the Backend! Instructing copilot to make full use of that during its autonomous development. Now you can just create an issue in GitHub (using GitHub MCP ofc :)), assign it to copilot and let it run. After it is done and tests are passing - review, remarks, done, deployed to Azure App Service!

Implementing efficient report building setup using Azure Durable functions and Stateful entities.
Utilizing Azure Storage and Azure Durable Task Scheduler backends
Building a system of reports building using Azure Durable orchestrations en Stateful entities.
The reports are composed of multiple, frequently reused parts.
Abstracting commonly used data preprocessing and aggregation parts as Stateful entities. Caching results to Blob storage.
Using stateful entities in orchestrations, effectively computing needed data only once for all reports. Achieving significant costs and time savings.

Software Engineer

Reveal

Building APE Multitenant, massively scalable, performant, K8s native, cloud agnostic service to process data
Solving complex problems of scaling, efficient parallelization, execution and providing a way to centrally manage processing of different kinds of data by simple, dedicated, pluggable, testable processing units. Allowing to focus and centralize corresponding processing expertise in simple units. 
Provides tools to implement, automatically test and ship processing units and their new versions.

Implementation:
Multiple services and client NuGet packages implemented in .net  
Kafka and RabbitMQ for work queuing, scheduling, distribution and result reporting
Platform provides exactly once semantics, client side exactly once is simplified using a client library.
Service is designed to run on any cloud platform and has built-in abstractions for long term cloud storage (aws s3, azure blob storage)

Setting up from scratch to a fully working Development process:
Aspire orchestration for local development and integrated testing with code coverage in GitHub actions runners hosted on k8s.
GitHub setup using self hosted GitHub Arc runners.
Enforcing process of PRs with required checks using GitHub rulesets:
- .net Code quality and smells testing. Majority are enforceable as compilation errors in VS, all of them are checked in a required GitHub action run.
- Unit and integration Aspire tests with differential code coverage reports for PRs.
- Helm chart to deploy the entire service, including 3rd party, to k8s.
- Automatically deploying the service to k8s and running integration tests for each PR as a required check.

Architect/Lead Developer/DevOps engineer

Continuation of the description above
Scaling:  
Using keda in k8s for scaling based on RabbitMQ throughput and mq length metrics.
Scaling to 1000s of pods to implement most compute hungry types of processing.

Testing:
Designing integration/api tests in a way to be able to run them in multiples modes: Aspire with code coverage and on a deployed service in k8s and report results to the PR
Automatically running Load tests for PRs in real k8s PR deployments
Automated Testing is arranged by labels and corresponding tests are executed in different contexts: Unit test, Aspire tests of the platform, Aspire test of processing units, full k8s integration tests of a real deployment per PR. 

Release:
Automated release process triggered by a GitHub release.  
Releasing tested helm chart of the platform and NuGet packages to be used by client applications to integrate to AWS CodeArtifacts

reveal – zylab

Based on success of Product Metrics Implementation for Legal Hold One a decision was made to implement the similar system for ZyLAB One set of products for purposes of functional monitoring, usage tracking and analysis. 

Designed model included ~400k metrics that were tracked over time for the entire SaaS operation.

Based on the metrics and time based queries the analysis were performed by Sales and Operations representatives 
A system of components including app, Opentelemetry collector was used to collect the metrics. 
To store and query metrics Prometheus was used.  PromQL ins and outs!
To visualize - Grafana. 

All this running on a single VM. All used products are opensource.

reveal - ipro

Designing and implementing distributed multi-tenant, multi-subscription, multi-region application - Legal Hold One service
Implementing all required infrastructure. Total infra Costs ~450 euro to host multiple deployments of the App and all related software
App: SQL store + Orleans middle tier + .net core API + Angular 9
CI/CD:
   Zero downtime deployment
   Fully automated deployment for all feature branches, staging, prod. Push & refresh - push to git refresh your browser windows
   Using Azure DevOps  to orchestrate CI/CD and run tests. (PowerShell/bash). Running and publishing results and code coverage of .net unit, .net integration, angular unit and e2e tests. Automatically making screenshots for failed e2e tests.
   Run-time: Azure Kubernetes service (AKS). Helm 3.0 to deploy. Using namespace security separation on both Dev, Staging and Prod. 
   Ultra cost efficient deployment: utilizing a single Dev AKS cluster for entire DEV process: Running Azure DevOps build agents for build, test and deployment orchestration, SQL server Dev as database and multiple app deployments( one per git feature/bug/release branch managed by Azure DevOps pipeline in automatic way). Scaling the infra down automatically based on demand, without affecting the app functionality. 
  Using Azure Managed SQL instance for Staging/Prod.
  Azure Application Insights for Front and Back end, Cost efficient setup, customizable filtering. 
  AKS monitoring plugin for AKS infra monitoring and alerting. Different settings for dev and prod. 
  Auto alerting on data from AppInsights and LogAnalytics.
 
Release management: 
After branching for release roll-out, Dev build testing is done. Next the build is automatically deployed to Staging on requested regions with a production DB restored. E2E test are executed on Staging. If everything is green last confirmation is received for PO and release is rolled out to Prod without affecting application users.

Architect/Lead Developer

Legal Hold One App: 
multi instance .net WEB API, Orleans for instance sync tasks, SQL server, EF core,  Automapper, JSON Patch for changes, Mvc.Testing for functional/integrations tests
.net Web Apis: Public(based on OData, used by clients) and Private(used by the Angular App)
Simple multitenant Admin portal ASP.NET MVC, internal
Rich Angular UI (the only thing I was only partially involved to)

Developing the app functionality and focusing on these non-functional parts of Legal Hold One service

Infra Automation: 
Automating multiple levels of infrastructure deployments using Terraform: azure/kubernetes/apps configuration/installation.  
Automatic secrets rotation  

Infra networking/security: 
Cilium OSS(Open source) implementation includes Hubble, l7 policies, all endpoints enforcement, host firewall, kube-proxy replacement with availability zone affinity support and denying/alerting on unexpected communication.
Kubernetes. Intrusion detection and notifications! With Cilium OSS!
Multizone traffic optimizations for Kuberentes

Observability:
Fully embracing OpenTelemetry for observability subsystems. 
Building product consumption tracking system based on top of OpenTelemetry metrics. 
PromQL ins and outs!

Multi-region Application deployment automation. 
Time to marked in multiple regions - 2 hours! With full cycle of automated testing(Unit/Api/Integration/E2E).

OAuth2 multitenant applications publisher verification: Azure Active Directory - MPN, Google.

Developing the app functionality and focusing on these non-functional parts of Legal Hold One

Implementing Monitoring/Auto-scaling setup in Azure Kubernetes Clusters using open-source software : 
Prometheus for metric collection 
Promtail + Grafana Loki for logs collection (Both index and data chunks stored in Azure Blob storage)

Grafana with AAD SSO + Access management for searching and visualizing Metrics and Logs

Alert-manager + slack/MS Teams to alert on Metrics and Logs

KEDA to auto-scale workloads based on various metrics from Prometheus

Implementing Access management: 
Users to the application: 
1. Azure Active directory + SSO +  Graph API
2. OKTA  + SSO
3. Short lived application secrets

DevOps engineers to AKS cluster:
Network access: Allowing from limited IPs
AAD Groups based authentication + Kubernetes RBAC
Monitoring tools access(Grafana) using the same AAD credentials

Automation(self hosted devops agents): 
In-Cluster: kubernetes service accounts + kubernetes RBAC
Azure resources:  Managed azure identities

Architect/Lead Developer/DevOps engineer 

Developing the app functionality and focusing on these non-functional parts of Legal Hold One

Automating infrastructure creation and updates using Terraform. 
Configuration split into few modules each creating bit of infrastructure. 
Using multiple end configurations - Dev and Prod.
Using multiple workspaces for Prod and Staging.
From zero to hero automation - from nothing to a running service(see App description bellow)

Deploying Azure resources (azurerm provider), deploying/configuring kubernetes resources using Helm, kubernetes and kubernetes-experimental Terraform providers.

Autoscaling various parts of deployments using Horizontal Pod autoscaler and AKS cluster autoscaler.
Using various metrics for scaling, including external metrics

Hardening security aspects of the deployment: 
 Implementing Web application firewall with full set of OWASP 3.2 rules (Opensource nginx + modsecurity) + Azure Log Analytics for logging and alerting for ALL environments (Automatically created branch environments as well!)
 Automatic pen-testing the app using OWASP ZAP proxy.
 Code security testing using SonarQube

Architect/DevOps engineer

Utilizing existing .net code on .NET Spark platform to do massive data analyses and enrichment
Utilizing Azure API management as a facade

Software Development Specialist

ZyLAB Technologies B.V.

Continuing integration of multiple Azure service into ZyLAB product portfolio.

Optimizing cost/availability/security ratios by using Azure Managed SQL offering with integrated security based on Azure identities. 
Implementing real-time reporting support for OLTP workloads using Columnstore indexes and Data partitioning

Azure WAF configuration, troubleshooting.

Adding SSO support for enterprise customers with Azure Active Directory.

Focusing more on technical side of the story.
Implementing graduate migration of the SaaS offering backbone to Microsoft Azure.
Implementing Cloud identity with combination of Azure B2C and B2B. ( Identity server for testing. ) 
SSO options for Client AAD tenants. 
Multi-Layered security model. Ability to use OpenIdConnect and Bearer tokens.
Developing Public Web API using Azure API management service 
Azure Application gateway with WAF in depth management and troubleshooting. 
Implementing monitoring and push notification for multiple Azure resources (AppInsights, Log Analytics).
Containerization(docker) of existing services with COM dependencies.
Secure infrastructure using gMSA and system managed service accounts. Using azure resource identities to implement secure access to Azure resources.
CI/CD extending Jenkins/azure Dev ops infrastructure and amount and types of tests including unit/web unit/smoke/integration tests.
Jenkins Multibranch pipeline implementation.
Interviewing candidates

Software and Solutions Architect

ZyLAB

Role with a mix of responsibilities: Architecture & Engineering, R&D management, Product ownership, Operations/implementations mentoring.

Managing a group of 20+ people.
 
Leading the migration to DevOps and implementing CD.

Sales support with Deep dive on site demonstrations and explanations to most important prospects. One of such resulted in the biggest deal in the history of the company.
 
Managing the biggest implementation of the product successfully.

Product release briefings and trainings.

Interviewing and on-boarding Software developers and QA engineers. 
Several hiring campaigns. Visiting Eastern Europe for face-to-face interviews.
Refining the team.

Leading high tech evolution utilizing various Machine Learning technics on Big data.
Continue enriching the feature set of the Product suit its performance, scalability and reliability.
Vastly improved automated testing infrastructure.
Unit tests are clustered into different groups and executed on a set of computers based on test external dependencies.
Different sorts of smoke-tests/integration tests automatically executed on different clusters of computers.
With heavily growing amount of tests integration time was reduced to the minimum without sacrificing quality by running only the tests that should be run for each integration cycle with a granularity of a source file change.
 
Reliability of the product improved vastly after allowing home grown Chaos Monkeys into the clusters while the tests are running.

R&D Manager/Architect

Leading a team of 8 developers
Continue improving Distributed Processing system - feature set, performance and scalability
Teaching company`s staff to use this configurable and powerful solution. 
Giving training sessions on system`s components, advanced configuration and extensibility. 

Working closely with operations to improve the system. 
Developed application that allows gathering of structural information from people in organization using fully customizable email templates and questioners.

Developed application to Review results of Processing. 
Features:
•	Near native representation of the document in your browser, no client install
•	Search on tags, metadata and content of documents.  
•	Automatic batches assign to reviewers. 
•	Tag/Bulk tag. 
•	Sampling. 
•	Redact documents. 
•	Produce a subset of the documents.
SQL server DB back end, Silverlight UI.

 Improving Production module.

Received best employees of the company award

Key achievements:
Complete application suite that takes unstructured mess as input, makes every piece of information (does not matter how deeply hidden) searchable. 
Close cooperation with operations

Architect

Leading a team of 5 developers. 
Implementing distributed system with single coordinator and multiple Processing nodes to perform EDiscovery related tasks
EF, SQL Server, MEF, All kinds of WCF (tcp-ip/http for node communication, Pipe for processing node to executing processes communication) 
Implementing WPF UI application to configure and monitor the system.
Setting up proper development environment CCNET, nant/msbuild build scripts, NUnit test execution on build server and automated Integration tests on cluster of computers.

Key Achievements
Product that was used right away by Operations and ZyLAB`s SAAS provider.
Good team, good progress.

Team Lead

Luxoft

Out-staffing. Client - Deutsche Bank
Working with DB development team on new applications portfolio using Silverlight and WPF technologies
Providing advices on best strategies an approaches. Implementing POCs to illustrate the core technical aspects of advice given
Member of Architects board in Deutsche Bank.

Have found time to improve working conditions of multiple people at DB ODC.
People in DB ODC in Kiev had no access to DB DNS servers. But there was a way to resolve addresses using a web page. I have implemented local DNS server that would inject itself on developer computer and resolve unknown DB addresses and cache them locally. After running the tool you could navigate the DB network like from inside. Many people in DB ODC have used the tool.

The Frayman Group

Leading a group of 5 developers 
Developing Rich Internet Application using latest Microsoft technologies (C#,  F#, Silverlight 3-4, WPF, Prism v2, Rx, ADO.NET data services, oData, WFC RIA services, WCF,  Entity Framework)
Applying expert system(JBoss Drools) to make decisions based on rules and know facts.
Integrating Apache Lucene for document search functionality.

Silverlight Architect/Team lead

OSUA Technology

Performance improvements of an 25+ years old C/C++ system      
Locating system's bottlenecks and eliminating them 
Applying different, much faster and scalable, algorithms for data manipulation.  (C, C++, gprof) 
Speeding up one of the most commonly used modules by a factor of 100 

Designing and implementing system wide Object Protection mechanism (C, C++)

Windows client for project management system. (C#, IronPython, WPF, GDI+): 
Dynamic content/Hosting IronPython in the client. Embedding dynamic WPF components animated by IronPython. (.NET Dynamic Language Runtime)
Build-in IronPython console  for real-time manipulation of application objects
Concurrent/background content layout calculations (Threading)
Presenting huge amount of data/dynamic content. 
Client-Server XML protocol analyzer (XPath) 
Extended text/xml analyses and import (XPath, Regular Expr )

Technical lead at OSUA Technology

Experienced software engineering leader passionate about building and scaling innovative solutions. I leverage deep expertise in software development, architecture, DevOps, and security engineering to help teams achieve business objectives, accelerate iterations, and deliver technically sound, results-driven products.

My approach fosters a clear, safe, and collaborative environment, empowering teams to experiment, move faster, and ultimately drive superior business outcomes. 

I thrive on continuous learning and possess the broad technical knowledge to confidently lead and execute the entire product lifecycle, from concept to successful operation.

I prioritize building solutions that are cloud-agnostic and minimize technological dependencies, focusing on open-source principles and ensuring adaptability for deployment on sovereign clouds with minimal modifications.

CKA: Certified Kubernetes Administrator (2025)
CKAD: Certified Kubernetes Application Developer (2023)
CKS: Certified Kubernetes Security Specialist (2024)
MC Azure Solutions Architect Expert (AZ-301, AZ-300)
MC Azure Security Engineer (AZ-500)
MC Azure Developer Associate (AZ-204)
MCSA: Web Applications 
MCSD: App Builder
MSCA: SQL Server
Certified Scrum Master.
Read team`s Trained Cobalt Strike operator

Taplio

Alexander Bartosh's Linkedin Analytics

Want to drive more opportunities from LinkedIn?

Taplio

.css-1jgreyo{background-image:linear-gradient(284deg, rgba(39,43,72,1) 0%, rgba(33,47,183,1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Alexander Bartosh's Linkedin Analytics

Want to drive more opportunities from LinkedIn?

Alexander Bartosh's Linkedin Analytics