I study how system ⚙️ behavior can be represented, maintained, and analyzed.
- 🧱 Runtime memory management & shared memory consistency across devices
- How computation stays correct across layers, devices, and schedules.
- 🔁 IR/Metadata based translation and pipeline-level understanding
- Understanding compute/graphic pipelines as state transitions, not code.
- ⚙️ Execution consistency analysis across system layers
- Seeing timing, ordering, and side-effects as structural invariants.
- 📝 Self-supervised learning for IR and 🌊 (dynamic)system-flow analysis
- Using ML to understand how a given system flows(runs) and evolve rather than existing representation.
I’m exploring how system-level execution (rendering, compute, drivers) forms graph-structured, particle-like dynamics:
- turning complex behaviors into DAG-like flow representations
- testing metamorphic stability, invariance, and convergence patterns
- treating IR + runtime patterns like a tiny dynamic system we can model
Ultimately, I’m trying to represent “how a system moves” —not just what it computes.