Особисті дані приховані Ветеран

Kyrylo Akimov
APPLIED AI ENGINEER / AGENT SYSTEMS
Ukraine | Remote | [відкрити контакти](див. вище в блоці «контактна інформація») | [відкрити контакти](див. вище в блоці «контактна інформація»)
[відкрити контакти](див. вище в блоці «контактна інформація»)

Applied AI practitioner with 12+ years in game development and hands-on work in autonomous agents, local LLM
workflows, retrieval and interactive prototypes. Focused on observable outcomes, evaluation and iterative system design.

12+ 2026 Local
Years in game development Independent AI systems R&D Models, tools & evaluation

01 SELECTED AI SYSTEMS
Trinity Sight | Autonomous Agent Harness | 2026 - present
Watch full demonstration | Local AI | RTX 3070 Laptop GPU, 8 GB VRAM
• Built a perception-action loop connecting environment observation, model-driven decisions, action execution and
outcome verification in Fallout 2.
• Combined local models, visual observations and available game-state data; recorded actions, model responses and
outcomes to investigate failures.
• Demonstrated autonomous completion of the Temple of Trials; the linked video shows one successful and one
unsuccessful attempt.
Tools: Python, local LLMs, LM Studio, computer vision, game-state integration.

Trinity Lablog | Local RAG Search & Evaluation
• Built multilingual notebook retrieval with Markdown chunking, local embeddings, incremental indexing and source
references.
• Compared embedding models on 23 lookup questions: relevant file in top 5 for 19/23 with Qwen3-Embedding vs 11/23
with the initial model; documented limits and negative results.
Tools: Python, NumPy, LM Studio, Qwen3-Embedding.

Local Voice Adventure | Voice Interaction & Evaluation
• Built a local Ukrainian voice interaction prototype with speech recognition, spoken feedback and response evaluation.
Explored continuous listening, interruptible playback and echo cancellation.
Tools: Python, HolosTTS, accentuk, FastConformer, Parakeet, local inference.

KYRYLO AKIMOV / APPLIED AI 1

Kyrylo Akimov
APPLIED AI / PROJECTS & EXPERIENCE

02 ADDITIONAL AI PROJECTS
Spark Harness | Local Coding Agent & Evaluation
• Built a local coding-agent and evaluation harness for Python tasks, including multi-file edits, test feedback and
independent checks.
• Used Docker to run generated-code tests in isolated, network-restricted containers; evaluated a local 4B model and
recorded successes and failure types.
• Evaluated a local Qwen3.5 4B coding agent on 18 diagnostic scenarios, three runs each: 10 of 42 task attempts
passed verification, and 12 of 12 negative-control attempts had the expected outcome. The set was not an
independent holdout.
Tools: Python, Docker, LM Studio, local LLMs.

KADO | AI Commerce Assistant
• Built an LLM-assisted product search prototype combining deterministic filters, lexical search, semantic retrieval and
optional local LLM query parsing.
• Added a Streamlit interface, FastAPI endpoint and evaluation cases for a synthetic catalog of 600 products.
Tools: Python, FastAPI, Streamlit, Qdrant, sentence-transformers.

Boxing Pose Experiment | Computer Vision
• Built a prototype to extract pose keypoints and motion features from boxing video; trained an experimental classifier
and evaluated it across separate recordings.
Tools: Python, pose estimation, feature extraction, machine learning.

03 PROFESSIONAL EXPERIENCE
AI Mentor | NeoVersity | 2026
Supported learners working with practical AI concepts and prototypes, including Python, retrieval and agent-based workflows.

Co-founder / Game Designer | Outta Range Games | 2023 - 2024
Led a Unity action roguelite from early prototype to playable demo; designed gameplay systems and coordinated up to 10
part-time contributors across disciplines.

Game Producer | Fortell Games | 2023 - 2024
Worked with the team to rework a Unity prototype, translating client direction into priorities and gameplay decisions.

Sound Designer / Lead Sound Designer | 2016 - 2024
Wargaming, Snowhound Games and Ink Stains Games. Audio design and implementation across shipped titles and live-game
pipelines using Wwise and Unreal Engine.

04 TECHNICAL SKILLS & LANGUAGES
AI & development: Python, local LLMs, agent workflows, retrieval / RAG, evaluation, FastAPI, Streamlit, LM Studio, Git, Docker
(test execution).
Interactive & production: Unity, Unreal Engine, Godot, ASR / TTS, Wwise, FMOD, Jira, Trello, Figma, Perforce.
Languages: Ukrainian and Russian (native); English (Intermediate, B1+).
Selected credits: Warhammer 40,000: Rogue Trader; S.T.A.L.K.E.R. 2: Heart of Chornobyl (trailer audio); Stoneshard; Deep
Sky Derelicts.

KYRYLO AKIMOV / APPLIED AI 2