Тимур

Tymur Katasonov
LinkedIn · Itch.io · GitHub · GitLab · Telegram
[відкрити контакти](див. вище в блоці «контактна інформація»)

Summary

• I am a Unity Developer with 8 months of commercial experience and 2 years of personal game
development experience. Comfortable across gameplay systems, async architecture, UI, and de-
veloper tooling. Interested in how AI can accelerate and improve the development process, and
have applied that practically, designing multiple project-specific agentic development pipelines.
Pick up new systems quickly and prefer understanding how things work over copying solutions.

TECH SKILLS & TOOLS
• C#, Unity
Animation, UI, TMP, AI, UniTask, NUnit, UPM, ShaderLab/HLSL, ScriptableObjects, Zenject,
VContainer
• Development and PM Tools
JetBrains Rider, Git, Jira, Slack, HacknPlan
• AI Development:
Claude Code, Ollama

CORE TECH & PROFESSIONAL SKILLS
• Knowledge of
OOP, SOLID, DI containers, Design Patterns, Async Programming, Clean Code principles
• Development Practices:
Debugging, Code Refactoring, Rapid Prototyping, Optimization
• AI & LLM Integration:
Prompt engineering, agentic pipelines, RAG, agentic retry pipelines
• Game Design:
I am passionate about and excel at designing deep and creative gameplay mechanics.
• Algorithms:
Classic on graphs, FSM, Behaviour Trees
• Soft Skills:
Strong communication skills, Team Leadership, Code Review, Agile/Scrum, Kanban
• Languages:
Ukrainian & Russian native
English C1 Advanced

WORK EXPERIENCE
• AI Solutions Lab LLC (Internship) October 2024 - May 2025
Unity Developer
DoodleDraw — a children’s iOS/Android app combining a multi-touch drawing board with
 ML-enhanced animations of the drawings, interactive story books with embedded mini-games.
Shipped to production and published on Play Market by a 9-person team, that included Unity
developers, artists and ML developers.

– Developed the responsive UI layout system supporting phone/tablet breakpoints via aspect
ratio detection, with live editor preview
– Implemented cross-platform permissions systems and local storage for settings
– Fully implemented one of the two interactive story books
– Built a full system for step-by-step tutorials
– Contributed to the Animator module, developed procedural skeleton animation for different
types of drawings

PROJECTS
• LegoKing GitLab Game Page Game Trailer
Turn-Based Tactics Game, 2-person team, March-June 2025
A Unity 6 turn-based tactics game with chess-inspired combat, a loot-driven equipment system,
and city-based progression across 10 battle levels.

– Designed a modular unit architecture with different unit classes, equippable weapons and
armor with synergies. This system incentivises the player to experiment with different unit
configurations and lets them build their own unique army, suitable for their playstyle
– Designed and developed the full battle rules engine as an async/await game loop. The rules
introduce unique battle concept, where player can only position the units, but they fight on
their own
– Implemented rotation-aware attack and movement patterns: killers move to the defeated unit’s
tile after each attack phase
– Built a data-driven SO-based system for easy configuration of units, items and scenes (HP,
damage, attack/movement patterns, class buffs, item models).
– Used Zenject for per-scene DI and a generic EventBus for zero-allocation cross-system com-
munication
– Built runtime factory that attaches player unit or AI unit controllers to the same prefab based
on team assignment

• ShaderForge AI GitHub
Unity Editor Plugin, Solo, March-May 2026
A Unity UPM package that generates HLSL/ShaderLab shaders from natural language descrip-
tions using LLMs (Unity 6). Generation combines URP specific templates, input expansion,
keyword-based knowledge base for RAG and auto-retry loop

– Designed a multi-provider LLM abstraction with 4 implementations: Claude API, OpenAI
API, a zero-cost Claude Code file bridge, and direct Claude Code CLI
– Built a full generation pipeline: user description → input expansion into a structured Shader-
Spec → knowledge base retrieval (190-effects) → template-based LLM generation → auto-retry
loop feeding Unity shader errors back into the prompt
– Built a 6-step Python data retrieval pipeline (synonym clustering, deduplication, taxonomy
schema) producing the knowledge base used for RAG
 – Developed a 5-mode A/B batch evaluation harness with CSV output to measure each pipeline
component’s contribution. Overall, success rate of shader generation went from 43% for raw
request generation up to 71% with full generation pipeline, which can be improved even fur-
ther by adding more data to the knowledge base

• DungeonLoop Game Page
2D Strategy / Tower Defense, Solo, 3 days, May 2026
A Unity 6 game where the player designs a dungeon to terrorize invading heroes and harvest a
fear-based resource.

– Built a runtime 8×8 grid with BFS pathfinding and a 5-state hero AI FSM using a weighted
room-scoring heuristic (fear, visit history, fog)
– Designed a fear economy with coroutine-driven Dread ticks, wave scaling, and a 5-wave Re-
solve debuff queue on hero flee
– Built a runtime dungeon editor: room placement, passage toggling, live cost display, BFS path
validation and a full snapshot/restore system

EDUCATION
• Bachelor of Science in Computer Science (2022-2026)
Taras Shevchenko National University of Kyiv, FCSC