What I'm focused on

Context: With limited DevOps resources, I took on cluster administration across prod/dev/stage.

What I’m doing:

• Setting up CI/CD pipelines for microservices
• Building ephemeral test environments for Bitrix24
• Taking Yandex Practicum course to systematize knowledge

Goal: Automate routine tasks and free up time for architecture.

Hypothesis: Running dev environments in Pods removes “works on my machine” issues.

Exploring:

• Yandex Spaces, GitVerse Codespaces, GitHub Codespaces
• Self-hosted setup on K8s + Theia/VSCode Server

Success metric: Developer starts coding in under 2 minutes without local setup.

Task: Semantic search across Confluence documentation with contextual answers.

Architecture:

• Ingestion: Airflow -> Confluence API -> Chunking -> Qdrant/ChromaDB
• Inference: Llama + RAG patterns + contextual ranking

Plan: Launch MVP with minimal team resources, gather quality metrics, prepare case study.

Focus: No-code/low-code tools for rapid AI integration into business processes.

Testing:

• Open WebUI + Ollama for local experiments
• n8n for orchestrating AI workflows
• Fine-tuning possibilities for corporate data

Deliverable: “Which tool for which task” checklist for the team.

Capabilities:

• Monitors site uptime and sends alerts
• Analyzes content changes and suggests updates
• Aggregates news by tags: K8s, Go, RAG, eBPF

Stack: Lightweight Python bot + cron jobs + vector DB for semantic search.

Context: Corporate pilot plus personal experiments.

Evaluation criteria:

• Code generation quality for Go/Python
• Integration with existing CI/CD
• Security and on-premise deployment options

Deliverable: Comparative matrix for adoption decision.

Challenge: Minimize content publishing effort for a non-technical user.

Workflow:

1. Tailor sends photo + description via Telegram
2. AI agent generates SEO text, tags, alt attributes
3. Preview -> approval -> deploy

Goal: Prove AI can bridge expertise and digital presence.

Philosophy: Every site feature is an experiment with minimal TCO.

Current experiments:

• Auto-generated OpenGraph images
• i18n routing without content duplication
• Git-tag deployments with automatic rollback

Rule: If a feature needs more than one hour of maintenance per week, it does not go to prod.

Focus: Systematizing knowledge in security and scalability.

Practical application:

• Modeling current projects in C4 to identify bottlenecks
• Adding security checklists to code review process

Deadline: Course completion - July 2026.

Approach:

• Time is blocked by themes, not tasks
• Weekly review: what gave energy, what drained it
• One focus rule: max three active directions at once

Tool: Simple Notion tracker and color-coded calendar.