About
I design number formats and RTL measured on a binary FPGA. GF-T started as a paper, became RTL that maps into Artix-7 fabric with no hard multipliers at all, and went through a SKY130 design submitted for fabrication — on a toolchain anyone can install for free. Before hardware I spent a decade building products and teaching: over a thousand developers, and a React Native course for the Russian-speaking internet.
Number formats designed against your constraints — ternary, low-precision, φ-based — each with an independent reference model and bit-exact vectors, not just a claim.
Synthesisable Verilog through an open flow: Yosys, nextpnr-xilinx, prjxray, iverilog. Measured on the binary FPGA ALINX AX7203 (Xilinx Artix-7 XC7A200T); a SKY130 design was submitted for fabrication.
Bit-exact conformance against models written from the specification rather than from the design — a check that can disagree with the RTL.
Over a thousand developers taught. Hardware is the current subject; the method has not changed.
A φ-based floating-point family, GF4 through GF1024, with the reference implementations that make it checkable.
A conformance catalogue with bit-exact vectors for FP8, BF16, MXFP4 and microscaling formats, published so anyone can validate their own arithmetic against it.
A camera bubble that floats above every other app and is never welded into the video. Screen, camera and pointer stay separate tracks until export, so the bubble can be moved, resized or reshaped after the take is over — and the glass rim refracts the screen behind it, because those pixels are still there. Native Swift on ScreenCaptureKit, AVFoundation and Core Image. The product page is at t27.ai/orb.
Based in Thailand, working with teams across Europe and North America.
Nothing I deliver requires you to buy a licence to reproduce it.
Contract and part-time hardware-AI, FPGA/RTL and ML-systems work.
The CV, the papers and the source are all one click away. If the work looks relevant, write — I answer.