CNN Market-State Directionality
Exploratory convolutional neural network framework for classifying market-state directionality from structured market inputs. Built to test whether localized price and volatility patterns carry directional information across regimes; shown as research workflow, not production alpha claim.
SOXX 3D Rolling Volatility Surface
Three-dimensional SOXX rolling volatility surface built to visualize regime shifts, clustering, and changes in volatility shape through time. Designed as a market-structure and risk-analysis tool, not a point-forecasting model.
Convolutional Neural Network (Adaptive Market State Strategy Testing)
I built a walk-forward market-state research project that compares CNN-based, statistical, and momentum-driven equity strategies over time. The findings suggest that return predictability is state-dependent rather than constant, with different strategies outperforming in different market regimes and nonlinear models adding value when feature interactions become more important. Built as a research and interpretation tool for comparative behavior across states, not as a live trading or return guarantee claim.
Photonic GPU Architecture
Built a GPU from scratch — custom 16-bit ISA, 9 SystemVerilog RTL modules, 4-warp 32-thread SIMD datapath, 5-stage pipeline, round-robin scheduler, banked register file (4×16×8×32b), Harvard memory architecture. Wrote a full compiler (lexer→parser→AST→IR→regalloc→codegen) targeting my own instruction set. Photonic accelerator subsystem with Clements MZI mesh, SVD optical matmul (W=UΣV†), 1310nm SiN waveguides, validated to 10⁻¹⁷ unitarity. All proofs pass. One command builds everything.
Project Five
Coming Soon
Coming soon.
Project Six
Coming Soon
Coming soon.