I build AI agents into the software development process. At ISD I’m rolling out an agentic SDLC on Claude Code, where an agent takes an issue all the way to a reviewed pull request. It rests on 14 years of production engineering, six of them as a team lead.

agentic-sdlc · one issue, end to endagent people
  1. Issueopened and assigned to the agent
  2. Reproducethe bug, or failing tests and invariants
  3. Planthen a critique of the plan
  4. Implementplus docs and simplification
  5. ReviewAI code review from three angles, plus an external AI reviewer such as Copilot
  6. QAand an ADR for the decision
  7. Pull requestengineers decide what goes in

Impacthow my reach grew

  1. Engineer2015–2020

    Senior on clinical software

    Grew to senior on the LIS genetics module and took on its most complex workflows. From 2017 on call for urgent customer issues.

  2. Team2020–2023

    A team through a stack change

    Became team lead in 2020. Moved the whole team from C#/.NET desktop to a Java/Angular web product; led my area on emergency support and client go-lives.

  3. Product2023–2025

    First client go-live

    Took our team’s part of the Results area live with its first client and fixed its performance bottlenecks, mostly in Oracle. Other LIS teams adopted my Grafana dashboards.

  4. Company2025–2026

    AI for the whole company

    Ran the company-wide intro to AI in development for 300+ people, in English, then moved to the AI team in June 2026 to build the agentic SDLC full time.

Experiencegit log --graph

89d9a50isd/ai-agentic-sdlcJun 2026 – nowCurrent

AI Engineer · ISD

  • Designing and rolling out an agentic SDLC at ISD on Claude Code, GitHub Spec Kit and GitHub Actions; in the pilot, the agent takes bug fixes from issue to pull request on its own.
  • Built its AI code review, used voluntarily by 50+ engineers on 8 repositories: 700+ pull requests reviewed, about two thirds of findings fixed.
  • Measuring the review continuously: ~98% precision (LLM-judged) and 52% recall on a control set of 25 known bugs.
  • Claude Code
  • Claude API
  • GitHub Spec Kit
  • GitHub Actions
  • MCP
5b6058cisd/lis-results-webMay 2022 – Jun 2026

Team Lead / Full Stack Developer · ISD

  • Led my team from the C#/.NET desktop genetics module to the Results area of LIS, a Java/Angular web product for lab results, and coached it through the new stack.
  • Grew the team from 8 to 13 (3 internal transfers, 2 external hires I interviewed); onboarded newcomers and documented onboarding; ran performance assessments and prepared promotion recommendations for management.
  • Planned and delivered releases for my area; took our team’s part of the product live with its first client; led my area on emergency support and client go-lives.
  • Fixed performance bottlenecks in 2023–2025, mostly in Oracle: execution plans and query optimization.
  • Built Grafana dashboards for the Results area that other LIS teams adopted.
  • Introduced GitHub Copilot in the team with the product’s first custom instructions.
  • In 2026 ran the company-wide intro to AI in development for 300+ people, in English.
  • Java
  • Spring Boot
  • Angular
  • TypeScript
  • Oracle
  • Grafana
d7099a2side/teachingFeb – Sep 2020

Course Instructor · Hillel IT School

  • Taught C#/.NET courses, from fundamentals to students’ first projects.
ae5dbd6isd/lis-genetics-dotnetOct 2015 – May 2022

.NET Developer → Team Lead · ISD

  • Built and maintained the LIS genetics module, a desktop app for genetic test workflows, results processing and reporting, where clinical data has to be exactly right.
  • Grew to senior level and became the team’s lead in 2020.
  • On emergency support on-call from 2017, 2–4 weeks a year, resolving urgent customer issues.
  • C#
  • .NET
  • WinForms
  • SQL
0c0dd11kzrk/erp-delphiJun 2012 – Oct 2015

Software Engineer · PJSC KZRK

  • Built and maintained the company’s in-house ERP in Delphi: budgeting, tenders, procurement.
  • Delphi
  • SQL

Projectshow they work

ISD · internal

Agentic SDLC

Autonomous bug fixes, with features in progress: open an issue, assign the agent, and it carries the work to a reviewed pull request.

Pipeline
Reproduce the bug, or write failing tests and invariants for a feature. Plan, then a critique of the plan. Implement, then a post-implementation slot (docs, simplification). AI code review plus an external AI reviewer, QA, and an ADR for the decision.
Built on
GitHub Spec Kit, Claude Code in GitHub Actions, GitHub Projects. Every step is templated and logged on the issue and the pull request.
Status
Pilot: bug fixes run end to end; features work with some limits.
  • Claude Code
  • GitHub Spec Kit
  • GitHub Actions
  • GitHub Projects
ISD · internal

AI code review

Comments on pull requests; engineers decide what to fix.

Three angles on code
An architecture overview; a focused review of risky spots found with a defect dictionary; and an unfocused whole-file read that catches what the focused pass was not told to look for, usually 1–2 more likely bugs.
Beyond the code
Checks changes against requirements and docs from the issue tracker over MCP. A second MCP server over an internal code graph maps the changed code to affected requirements and test cases; the model filters them into a regression checklist.
How it’s measured
An LLM job runs three times a day: it checks which findings were fixed, judges false positives and tracks adoption. Recall is checked separately on a control set of known bugs.
Planned
Develop the review through the agentic SDLC: open issues automatically from review logs and metrics (missed and wrong findings) and let the pipeline fix them, so the review maintains itself for obvious defects and simple features.
  • Claude Code
  • MCP
  • Code graph
  • Multi-agent
personal

claude-mentor

A custom Claude command for learning any topic that keeps your progress between sessions.

  • Claude Code

Skillsby area

AI engineering

Agentic SDLC, Claude Code, Claude API, MCP, GitHub Spec Kit, multi-agent review, context engineering, LLM evaluation

Backend

Java, Spring Boot, C#, .NET, REST APIs

Frontend & desktop

Angular, TypeScript, WinForms, Delphi

Data & performance

Oracle (execution plans, query tuning), SQL, T-SQL

Infrastructure

GitHub Actions, Docker, Prometheus, Grafana

Leadership

Hiring and onboarding, performance reviews, release planning, mentoring, technical training and talks, incident response

Also worked with: Kubernetes, Helm

Certifications

Languages & education

  • Kryvyi Rih Technical University · Master’s, Software Engineering2013
  • UkrainianNative
  • EnglishC1 · Advanced
© 2026 Hryhorii Manaienkovhman.pro