MishaMechina
Simulation engines, data platforms, and AI-augmented tools—built on hard data, shipped to production.
Systems
Working builds, experiments, and prototypes. Each includes stack + architecture.
Auratio
Precious metals portfolio platform with AI-powered insights and compliance-first governance.
- →LLM scenario explanations with compliance gating
- →Market brief generation and premium trend narration
- ✓Live at auratio.app — web and mobile (iOS/Android) with shared pricing API
- ✓D3 + Recharts visualisations for ratio analysis, premium trends, and landed cost modelling
Tactica Engine
Full-stack battle simulation and strategic analysis platform with a three-mode interface.
- →Rules infrastructure ready for ML/NN integration
- →R-trained Random Forest models for outcome prediction
- ✓Three-mode UX: Tactical Recon (list building), Battle Analysis (matchups), God Mode (full sim control + AI goals)
- ✓LLM orchestration: AI generates analysis specs → backend runs batches → AI summarises findings
Adeptus-Scraper
Deterministic rules engine and simulation kernel. Backend service — no public UI. Powers Tactica Engine.
- →Sim kernel is AI-free by design — deterministic for clean training data
- →Optional LLM PDF extraction layer (under review)
- ✓384 tests, 0 failures — seed-based reproducibility guarantees identical outcomes per seed
- ✓96.5% data coverage: 1,292 datasheets, 3,652 weapons, 1,234 stratagems
Systems grounded in reality
I build systems grounded in reality—where AI works with data, not guesswork.
My approach comes from operating in real environments, not just designing for them. I focus on platforms that produce outputs you can test, audit, and rely on—whether that's automation, reporting, or simulation.
AI, to me, is a partner. It should be allowed to explore beyond human assumptions, but never operate without structure or accountability. The balance is simple: give it freedom to find better paths, then validate everything against real data. Trust—but verify.
I work across the full stack to ship systems that exist today, not just roadmaps for tomorrow. AI extends what already works—it doesn't replace the need for solid logic, clear constraints, and human responsibility. Because in the end, the risk isn't AI—it's how people choose to use it.
- LanguagesTS, Python, R, SQL
- FrontendNext.js, React Native
- InfrastructureFirebase, Vercel, GCP, Docker
- AI/MLLLMs, Random Forest, FastAPI
Let's connect
Collaboration, technical discussion, or interesting problems.