jordan@waterloo:~$ cat mobile.txt
mobile version: work in progress.
open this on a desktop for the full site.
jordan@waterloo:~$
continue to the desktop anywayJORDAN-OS [FAILED] Desktop recursion limit exceeded. Kernel panic: recursive display server. Restarting display manager…
~/hardware/ # where I am now mvm_accelerator/ 128 lanes, 450 MHz, ~850 GOPS tanh_pipeline/ 280 -> 840 MHz over four rewrites wordle_on_kria/ FSM in fabric, baremetal C on the PS morse_code_riscv/ ASCII to Morse on LEDs, bare-metal RISC-V asm reflex_meter_riscv/ 0.1 ms reaction timer, polled, on a HiFive1 ~/software/ # things I built because I wanted them to exist summarization_tool/ doc ingestion to LLM eval, in testing at Health Canada uw_awards_search/ searchable mirror of the UW Awards Directory sred_copilot/ T661 drafts from markdown evidence, with citations linxicon_solver/ optimal word chains, daily replay at /linxicon-solver/ ~/papers/ # where I came from microgrid_rl/ first-author, CUCAI 2026, -23% diesel rag_agents/ co-author, CUCAI 2025
3B Electrical Engineering at the University of Waterloo. I started in hardware: PCB design on the Midnight Sun solar car, then boards through EMI/EMC compliance at Harding Technologies. Then I went up the stack.
From September 2024 to April 2026 I was at Wat.AI, leading an 11-person team on a reinforcement learning controller for off-grid microgrids in rural sub-Saharan Africa. It worked: 23% less diesel per year at equal reliability, and a first-author paper at CUCAI 2026.
Then we tried to deploy it. The low-cost board we needed had doubled in price and was out of stock, squeezed by the AI compute shortage. The inference hardware cost more than parts of the microgrid it was supposed to control. The algorithm was never the problem. The hardware was.
So I went back down. ECE 124, 222 and 327 done; the MVM and tanh accelerators in ~/hardware came out of that, timing closed through architecture instead of tool flags. This term I'm writing SystemVerilog for the Zynq UltraScale+ on my own time. Along the way: ML Engineer at Health Canada's PMRA, where the summarization tool in ~/software is now in testing with toxicologists, and an experiment at the National Research Council I designed to quantify volumetric airflow from flat-roof vents, with about $100,000 of lab equipment specced for it.
Winter 2027: Asia, on the Cansbridge Fellowship. Fall 2027: open.
2026 May–Aug VP Services, Waterloo Engineering Society 2026 Jan–Aug ML Engineer (Research), Health Canada (PMRA) 2026 Jan–Apr Director, Wat.AI 2025 May–2026 Apr Technical Project Manager, Wat.AI 2025 May–Aug Research Assistant, National Research Council of Canada 2024 Sep–2025 Apr Core Member, Wat.AI 2024 Sep–Dec Teaching Assistant (ECE 150), University of Waterloo 2024 Jan–Apr Engineering Co-op, Harding Technologies 2023 Sep–Dec Hardware Designer, Midnight Sun Solar Rayce Car
{
"looking_for": [
"FPGA / RTL co-op. Winter 2027 in Asia, Fall 2027 anywhere.",
"Winter is the Cansbridge Fellowship term, so Asia is fixed. The rest is open.",
"If you do ASIC design or manufacturing there, I want to talk.",
"If you have been through a full tapeout, tell me how IP licensing and",
"mask costs change the decisions I make in RTL.",
"",
"Hardware is first on the list, not the whole list. ML systems and the",
"software between the model and the silicon count too."
],
"email": "jordan.jay.leis@gmail.com",
"linkedin": "https://www.linkedin.com/in/jordan-leis/",
"github": "https://github.com/Jordan-Leis",
"resume": "https://drive.google.com/file/d/1jHbSPjZX3hLAVGLmC30Kj_MwDdtXkR9Z/view?usp=sharing"
}
The site may not allow itself to be shown inside another page, or it's taking too long.
Open in your browser ↗