::engineer & builder — Guelph, Ontario

Mani Rash Ahmadi

i make things. paintings, beats, robots — lately, the tools that make chips.

my goal is technology that lets anyone make anything. right now that means closing the loop between machine learning and silicon.

::Currently

first principles research — consulting swe
flodraulic systems — systems engineer
uog gharabaghi group — research engineer

fig. 01 — ascii animation · 80×40 chars · 262 frames

~11^min pytorch model to silicon PPA metrics, end to end

85⁺ binary record types reverse-engineered, zero docs

2,000⁺ spectre simulations orchestrated by CMA-ES

1–15^% of published ibex PPA refs across 12 configs

01Now

finishing my engineering degree at the university of guelph
building analog-gradients — LLM agents orchestrating commercial EDA tools: PyTorch model in, silicon PPA metrics out, ~11 minutes end to end
reverse-engineering undocumented industrial binary formats at flodraulic systems
running calibur labs — the long game is compute sovereignty: chip design accessible enough that intelligence infrastructure stays distributed

02Selected work

2025 —

analog-gradientsflagship

autonomous chip design exploration. produced a working 4-lane SIMD GPU and ran the lowRISC Ibex core through 12 configurations within 1–15% of published references.

4-lane SIMD GPU · 25.8 kGE · 118 MHz · 0.6 mW — from a pytorch model 2026