Closed-loop in-memory computing
Physical systems for matrix equations, inverse problems, and continuous-time computation using SRAM and emerging memory technologies.
Analog computing · AI hardware
Piergiulio
Mannocci
I research analog and in-memory computing systems that use circuit dynamics to solve problems in artificial intelligence, scientific computing, and optimization.
Latest
Recent publications, preprints, software releases, talks, and professional updates.
Aug 2026
Publication
Highly-Uniform Passive Crossbar Arrays of Resistive Switching Random Access Memory (RRAM) for In-Memory Computing Applications
Jul 2026
Software
32 × 32 passive RRAM crossbars with 99.5% functionality and <3% programming error, demonstrated across image processing, classification, combinatorial optimization, and nonlinear regression.
loopus² v1.0 released
Jul 2026
Career
Open-source Python framework for behavioral simulation and architectural benchmarking of analog in-memory computing circuits.
Joined Unconventional AI as a Member of Technical Staff
Jul 2026
Preprint
Working on system modeling and architectures for energy-efficient AI hardware.
Continuous-time nonlinear closed-loop in-memory computing for high-accuracy massive MIMO detection
Jun 2026
Preprint
A closed-loop analog solver with nonlinear processing and iterative refinement for high-order MIMO detection.
Modern analog computing for solving differential and matrix equations
Jan 2026
Publication
A perspective on continuous-time analog systems for scientific computing and linear algebra.
Fully integrated analogue closed-loop in-memory computing accelerator based on SRAM
Published in Nature Electronics.
Research
My work sits at the intersection of circuits, dynamical systems, emerging memories, and hardware–algorithm co-design.
Analog optimization and communications
Nonlinear dynamical circuits for massive MIMO, compressed sensing, constrained optimization, and scientific computing.
Reliable AI hardware
Architecture, modeling, precision enhancement, and hardware–algorithm co-design across analog and mixed-signal computing systems.
Research software
Open tools developed to support reproducible modeling and evaluation of analog computing systems.
Open-source · Python · v1.0
loopus²
A behavioral simulation and architectural benchmarking framework for open- and closed-loop analog in-memory computing circuits.
loopus² combines compact models of memory arrays, operational amplifiers, data converters, and programming circuits with interchangeable analytical and LTspice simulation backends.
- Operating-point, transient, nonlinear, and noise analyses
- Crossbar nonidealities and programming variability
- Area, latency, energy, power, and performance benchmarking
Featured work
A selected set of publications representing my current research direction and earlier work.
Nature Electronics · 2026
A fully integrated analogue closed-loop in-memory computing accelerator based on static random-access memory
First fully integrated SRAM-based closed-loop analog in-memory computing accelerator for directly solving linear algebra problems.
Advanced Electronic Materials · 2026
Highly-Uniform Passive Crossbar Arrays of Resistive Switching Random Access Memory (RRAM) for In-Memory Computing Applications
Highly uniform 32 × 32 passive RRAM arrays with precise multilevel programming, experimentally validated across image processing, neural classification, combinatorial optimization, and nonlinear regression.
arXiv preprint · 2026
Continuous-time nonlinear closed-loop in-memory computing for high-accuracy massive MIMO detection
Nonlinear closed-loop analog processing combined with iterative refinement for accurate and energy-efficient massive MIMO detection.
Science Advances · 2025
Seizure detection via reservoir computing in MoS2-based charge trap memory devices
A neuromorphic reservoir-computing system based on emerging memory devices for biomedical seizure detection.
Advanced Materials · 2023
Reservoir computing with charge-trap memory based on a MoS2 channel for neuromorphic engineering
Demonstration of neuromorphic reservoir computing using the nonlinear dynamics of MoS2-based charge-trap memory devices.
About
Computing through physical dynamics.
I work on analog and in-memory computing architectures that use circuit dynamics to solve linear algebra and optimization problems directly in hardware. My research spans mixed-signal circuit design, emerging memories, dynamical systems, and hardware–algorithm co-design.
I am currently a Member of Technical Staff at Unconventional AI. Previously, I was an Assistant Professor at Politecnico di Milano, where I worked on closed-loop computing systems based on SRAM, RRAM, PCM, and other emerging technologies.
Selected talks
Invited presentations on analog and in-memory computing.
2026
Solving inverse problems with in-memory computing: from linear algebra to optimization
International Workshop on Ising Machines, Singapore
2025
Towards robust, scalable, and deployable analog in-memory computing with resistive memories
IEEE International Conference on Emerging Electronics, Bengaluru
2025
Recent advances in analog in-memory computing with resistive memories
Neuromorphic Symposium, Paris