S

SEMRON Tech Stack

AI chip design with analog in-memory compute architecture

Semiconductor Manufacturing Dresden 11–50 employees Privately Held

SEMRON designs custom AI chips optimized for inference density, built around analog in-memory compute rather than conventional digital architectures. The stack — Verilog, SystemVerilog, Spectre, TCAD simulation, PyTorch, and a custom compiler toolchain (LLVM/MLIR/IREE) — reflects a hardware-first design philosophy where chip architects and compiler engineers work tightly coupled. Core pain points cluster around quantization methods for analog hardware and compiler scalability, indicating the company is still solving fundamental algorithmic problems at the chip-compiler boundary, not yet shipping at volume.

Tech Stack 14 technologies

Core StackPython MATLAB C++ PyTorch Verilog-A Spectre DATEV LabVIEW LLVM MLIR IREE C/C++ Verilog SystemVerilog

What SEMRON Is Building

◆Challenges

Optimierung interner prozesse
Klärung buchhalterischer fragestellungen
Scaling compiler toolchain
Efficient high-accuracy inference on analog in-memory compute platform
Quantization methods tailored to in-memory compute constraints
Running models efficiently on analog compute platform
Geo-distributed qat framework
Quantizing models for analog hardware

▲Active Projects

Design analog-aware quantization methods for in-memory compute platform
Tcad simulation
Verilog-a & spectre modeling
Process engineering for device fabrication
Design mechatronic assemblies for automated test & characterization of dies/wafers of custom ics
Program automation routines for automated testing & characterization
Design pcbs for prototype systems with our custom chips
Scalable compiler toolchain for future chip generations
Compiler operations and optimization passes
Generic nn compilation for semron hardware

Hiring Activity

Decelerating10 roles · 2 in 30d

Department

Engineering

6

Research

2

Finance

1

Ops

1

Seniority

Senior

5

Mid

3

Intern

1

Junior

1

Company intelligence

Find more companies like SEMRON by tech stack, pain points and active projects

Get started free

About SEMRON

SEMRON is a 11–50 person semiconductor design company based in Dresden, Germany. They specialize in custom AI inference chips built on analog in-memory compute principles, a departure from standard digital GPU/TPU approaches. The team is structured as engineering-heavy (6 engineers) with dedicated research (2), finance, and operations; hiring has decelerated recently with only 2 roles posted in the last 30 days. Active work spans analog-aware quantization, TCAD device simulation, custom IC layout and testing automation, and a compiler toolchain to map neural networks onto their proprietary hardware. The company operates across the full stack: from silicon process engineering and device modeling through fabrication test characterization, prototype PCBs, and software compilation.

HeadquartersDresden

Company Size11–50 employees

Frequently Asked Questions

What tech stack does SEMRON use?

Hardware design: Verilog, SystemVerilog, Spectre, Verilog-A, TCAD. ML/compiler: PyTorch, LLVM, MLIR, IREE. Supporting: Python, C++, MATLAB, LabVIEW, DATEV.

What is SEMRON working on?

Analog-aware quantization for in-memory compute, TCAD simulation, custom IC design and test automation, compiler toolchain for neural network mapping, and PCB design for prototype systems.