Simra AI
EDA orchestration for analog and mixed-signal teams: existing tools, private environments, repeatable evidence, and engineer signoff. Simulation and debug loops, PDK-aware migration checks, verification evidence, workflow memory.
Dr.-Ing. · Research Engineer
AI agents, RL environments, and evaluations for chip design and EDA workflows.
Co-founder of Simra AI. Before that: four tapeouts as a chip designer, from X-FAB 130 nm energy harvesting to 22 nm FDSOI Ethernet PHY circuits.
Expertise Analog and mixed-signal — the corner of chip design automation left behind. Digital is synthesised and largely automated; analog is still hand-drawn and simulation-bound, roughly 40× less automated. That gap is the work.
Autonomous chip design house — a team of agents designing a chip toward tapeout, streaming 24/7
I spent years doing chip design the hard way: schematics, layout, simulation, post-layout debug, tapeout pressure, and the small judgement calls that never fit cleanly into a textbook. Simra is the tool I wanted during those loops.
Within chip design, my expertise is analog and mixed-signal — and that is deliberate. Digital logic has three decades of synthesis, place-and-route, and formal tooling behind it; a designer there writes intent and lets the flow build the circuit. Analog has almost none of that. Devices are still sized by hand, layout drawn by hand, and correctness established one long simulation at a time. The result is a design-effort gap of roughly 40× against digital — and it is the part of the industry where good agents have the most left to give.
So my work spans EDA orchestration and the creation of RL tasks, evaluations, and benchmarks for engineering agents, aimed at exactly those loops. Simra sits beside existing toolchains, runs repetitive work, captures evidence, and brings engineers back in when judgement matters.
Before Simra I explored data ownership, crypto, agents, and consumer products. A lot of that was useful training, but the real pull was back to chip design: a difficult, high-trust field where better software can save expert engineers from endless manual glue work.
EDA orchestration for analog and mixed-signal teams: existing tools, private environments, repeatable evidence, and engineer signoff. Simulation and debug loops, PDK-aware migration checks, verification evidence, workflow memory.
A 24/7 team of autonomous agents designing a chip toward tapeout, streamed live on YouTube and Twitch.
IMTEK, University of Freiburg. Multidisciplinary systems engineering across mechanical, fluidic, and electrical domains, with a focus on circuit and system design. First tapeout here: a fully integrated charge pump for energy harvesting in X-FAB 130 nm, ref [3].
PhD at TU Berlin with multiple tapeouts, including 22 nm FDSOI high-speed Ethernet PHY circuits and source-terminated DAC driver work. Publications below, refs [1]–[8].
Taught electromagnetism, then built Datalatte and survey/data experiments. The useful lesson: start from a painful workflow, not from an ideology about technology.
Hackathons as quick product reps around privacy, data, social graphs, and agents: ETHMunich (2nd place, AI track), ETHRome (5 bounties), ETHGlobal Bangkok (Circles 1st bounty), and the SAFE Agentathon (agent matchmaking).
Co-founder. Agents, RL tasks, evaluations, and benchmarks for chip design and EDA workflows, plus the live autonomous design house.