https://github.com/dontkaiad/it-journey
a structured overview of my career in tech: from L1 support to LLMs (a bit of chaos included)
https://github.com/dontkaiad/it-journey
api confluence documentation grafana jira linux monitoring prompt-engineering qa rpi sql ssh tech-support
Last synced: about 1 month ago
JSON representation
a structured overview of my career in tech: from L1 support to LLMs (a bit of chaos included)
- Host: GitHub
- URL: https://github.com/dontkaiad/it-journey
- Owner: dontkaiad
- License: other
- Created: 2025-07-20T08:24:12.000Z (11 months ago)
- Default Branch: main
- Last Pushed: 2025-11-14T08:40:07.000Z (8 months ago)
- Last Synced: 2025-11-14T10:21:35.818Z (8 months ago)
- Topics: api, confluence, documentation, grafana, jira, linux, monitoring, prompt-engineering, qa, rpi, sql, ssh, tech-support
- Homepage:
- Size: 52.7 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
- Support: support-templates
Awesome Lists containing this project
README
# it-journey
a structured overview of my career in tech: from L1 support to LLMs (a bit of chaos included)
---
## 👋 Who I am
Hi, I’m **Karina Lark** (aka **Kai**).
I’m a tech generalist with a background in **technical support, QA, monitoring, and LLM workflows**.
Neurodivergent. Driven. Curious. Creative.
From hardware to prompts, from chaos to clarity.
---
## 🧩 What’s inside this repo
This portfolio contains selected examples from my work in:
- **L1–L2 technical support** (Avito, Joom, Zvuk B2B, etc.)
- **CI/Monitoring tasks** (Grafana, Jenkins, Graylog, webhooks)
- **QA & testing** (manual, smoke/regression, test cases)
- **LLM prompt design** (Wikibot system prompts, few-shot logic)
- **Documentation** (internal knowledge bases, Confluence)
---
## 🛠️ Stack & Tools
| Domain | Tools |
|--------|-------|
| Monitoring & Logs | Grafana, Graylog, OnCall |
| Automation | Jenkins, Gitlab, webhooks |
| QA | Jira, test-cases, bugtracking |
| Docs | Confluence, Markdown |
| Tech Support | UseDesk, RDP, SSH, REST API |
| OS | Windows, Ubuntu, macOS |
| Languages | Python (basic), SQL, Bash |
| Prompting | LLMs (GPT-based), few-shot, system-level |
---
## 🧠 About me
- 💻 5 years in tech support (B2C → B2B → edge devices)
- 🔁 Cross-functional: worked with DevOps, QA, playback, hardware
- 📊 Proactive: traced incidents, helped to build monitoring, optimized workflows
- 🧪 Curious: learned QA by testing edge devices hands-on
- 🤖 Explored LLM prompting before it was cool.
me return manifesto
> me return 422 when no one knows what the bug is
> me return 504 when devs ghost me
> me return 418 when asked to write a press release
---
## 🐾 Fun facts
I also:
- play League of Legends (10k hours as adc main, MT peak - easter egg in the end of README)
- using urbex as my cardio
- write postpunk music and raise three cats.
---
## 📫 Let’s connect
Feel free to message me:
📧 icymirum@gmail.com
💬 [Telegram](https://t.me/hey_lark)
🟢 [WhatsApp](https://wa.me/79118145193)
🐙 [GitHub](https://github.com/dontkaiad)
📍 St. Petersburg, open to relocation (PL, RS, MNE pref.)
---
🕹️ Easter Egg: Champion Spotlight — Kai, the Neurodivergent Engineer
**Role:** Marksman / Support Hybrid
**Resource:** Focus (instead of Mana)
**Damage Type:** Mental / True
**Scaling:** Knowledge + Concentration
**Origin:** Tech Support Depths
**Patch:** Pre-PromptOps
---
### 🧬 Passive — *Overfocus*
Kai accumulates Focus while listening to russian post-punk playlists and deep-diving into chaotic systems, log clusters, or half-written specs.
Focus builds faster when the task is obscure, undocumented, or happening under pressure.
Once Focus reaches 100%, her next ability is Overloaded, granting enhanced insight or performance.
*⚠️ Staying in Overfocus mode too long triggers Mental Overload — abilities are weakened and Kai becomes distracted, muttering about dashboard inconsistencies.*
To reset Overload, Kai must briefly switch context (e.g., take a short offline break, look out the window, touch grass, or reheat coffee).
This clears the stack and restores baseline clarity.
**Bonus:**
If Kai opens a can of Monster Energy™ during work — Focus gain is doubled for 10 minutes.
---
### 💡 Q — *Pattern Lock*
Kai highlights invisible structures within chaos: API chains, failing workflows, undocumented triggers.
Perfect for log diving, error tracing, or identifying mismatched assumptions in client complaints.
**Bonus:**
Holding Monster increases speed of activation.
*• Enhanced: auto-links incidents to root cause and suggests Confluence anchor.*
---
### 🧪 W — *Debug Dance*
Executes manual probing across systems using smoke tests, user scenarios, and synthetic prompts.
Each cast increases attention and speeds up bug hypothesis formation.
*• Stackable. At 3 charges, enters "Flow State" — reducing cooldowns and increasing insight radius.*
*• Each consecutive cast increases chance of Overload buildup.*
---
### 🧠 E — *Prompt Mirage*
A long-cooldown ability. Not spammable.
Effective only if Kai is at moderate Focus and no context-switch occurred in last 3 minutes.
Summons a well-structured prompt tailored to trick even the most stubborn LLMs.
Applies few-shot / zero-shot or CoT logic depending on context size.
Enemies hallucinate bugs. Allies call it wizardry.
*• Can backfire beautifully if model isn’t aligned.*
*• Enhanced: doubles accuracy of bot-assisted support.*
*• Overloaded version: generates two alternative prompt strategies, but at risk of overthinking.*
---
### 🎧 R — *Reality Collapse* (Ultimate — unlocks at 6 / 11 / 16)
Kai collapses surrounding ambiguity into structure.
Documentation writes itself. Support workflows optimize. Focus spikes.
At:
- **Level 6**: Analyzes unknown systems with limited data.
- **Level 11**: Connects edge cases to high-level patterns.
- **Level 16**: Sees the architecture behind incidents — and updates the Wiki.
*• Passive bonus: musical damage over time (if headphones are on). Press ctrl5 to put headphones on.*
---
### 🐾 Champion quotes:
> "I don’t scale. I snowball."
> "It’s not chaos — it’s undocumented structure."
> "Stacking Flow. Don’t ping."
> "One cat runs my cluster. The other manages incident tags."
> "I don’t wait for permission to carry."