Curitiba, Brazil (Remote) | English (Fluent) | Portuguese (Native)

Email: alisonskrinski@gmail.com

Quick Links
Resume | LinkedIn | LoRA Evaluation Project | Forgotten Eras on Steam

I design systems with a strong sense for structure, hidden constraints, and downstream consequences. I am good at understanding what someone is actually asking for, stress-testing the assumptions inside that request, and turning vague requirements into something concrete, logical, and workable.

I care about rigor, clarity, and good coding practices. I like explicit logic, maintainable structure, and systems that hold up when reality gets messy. AI-assisted development is part of my workflow, but the value I bring is judgment: taste, far-sight, problem framing, and the ability to carry an idea from a fuzzy need to a real implementation. I like work that starts messy and becomes clear once the right system exists.

Most of my work fits around internal tools, workflow infrastructure, evaluation systems, automation, and system-heavy feature work that needs both design sense and technical discipline. I am especially strong in architecture-heavy work where someone needs clear logic, strong boundaries, operator-friendly tooling, and a practical path from requirement to implementation.

• I can ingest requirements quickly, spot bad assumptions early, and see where a design will break before it does.
• I think in systems: dependencies, data flow, edge cases, operator clarity, and long-term maintainability.
• I am rigorous about logic, structure, and code quality.
• I use AI for leverage, but the human value is still mine: judgment, direction, taste, and problem-solving.

(August 2024 - Present)

Key contributor on the playable demo for Forgotten Eras.

• Designed and implemented modular gameplay systems, interaction logic, and support tooling in Unreal Engine.
• Translated design needs into clear, maintainable system structures using Interfaces, Actor Components, and reusable logic.
• Caught gameplay softlocks, pipeline failures, and implementation risks early enough to keep the build stable.
• Worked remotely with documentation, task tracking, technical ownership, and cross-discipline communication.

(June 2019 - Present)

• Worked across 3D pipelines, technical planning, modding, reverse-engineering, and system-heavy personal projects.
• Built a habit of breaking ideas down into requirements, limitations, object relationships, and implementation risk before building.
• Solved problems in messy environments where tooling, documentation, or support were limited.

LoRA Evaluation Project
An end-to-end system I authored for inspecting how Low-Rank Adaptation (LoRA) models affect diffusion-based image-generation workflows.

I built this after noticing a real problem in image-generation AI: most LoRA evaluation is still done by vibes. People generate a few images, guess what changed, and hope the answer is meaningful. I wanted a system that could store evidence, compare controlled runs, and make the behavior inspectable later.

That meant learning the domain well enough to work with LoRAs, diffusion workflows, ComfyUI, and the practical realities of evaluating deep-learning systems without pretending to be doing AI research.

Baseline vs LoRA comparison inside the review surface.

• Designed the architecture around extraction, validation, persistence, review assembly, and operator-facing presentation.
• Built a pipeline spanning ComfyUI ingestion, validation gates, SQLite-backed evidence storage, diagnostics, and a local review app.
• Treated the whole thing as an instrument: explicit boundaries, inspectable evidence, reproducibility, and truth over guesswork.
• Took it from the original problem all the way to a working, usable system authored end to end by me.

This is the strongest proof of how I work: I see the design problem, I learn what I need to learn, I think through the edge cases, and I build the system that solves it.

Before this project, I was already solving the same class of problems inside game development: designing systems, building tools, structuring logic, and making workflows easier to operate.

• Modular Building System: state-driven placement workflow with validation, mode handling, and undo/redo logic.
• Universal Camera Manager: reusable camera behavior with predictive occlusion handling, smoothing, and modular control.
• Procedural Spline Tools: editor tooling for faster environment assembly and less repetitive setup work.
• Tactical AI Flanking System: EQS and Behavior Tree logic for dynamic positioning, line-of-sight, and decision support.

Different domain, same strengths: system design, logic, tooling, problem anticipation, and implementation that stays readable under pressure. I am strongest when I can bridge design intent with implementation reality and turn a messy requirement into a practical system people can actually use.

Python TypeScript React FastAPI SQLite Git Unreal Engine 5 Blueprints UMG Niagara Chaos Blender ComfyUI diffusion workflows LoRA evaluation

If you want to talk about systems work, AI-assisted software development, tooling, or a role where I can help turn messy requirements into working software, reach out:

Email: alisonskrinski@gmail.com
LinkedIn: linkedin.com/in/alison-krinski