Eye On A.I. – Ep. 329: Izhar Medalsy – How AI Solves Quantum Computing’s Biggest Problem

Host: Craig S. Smith | Guest: Dr. Izhar Medalsy | Date: March 31, 2026

Episode Overview

This episode dives into the intersection of artificial intelligence and quantum computing with Dr. Izhar Medalsy, co-founder of Quantum Elements. The conversation focuses on how AI-powered digital twins—high-fidelity simulations of quantum hardware—are making noisy, error-prone quantum systems more useful and paving the way to practical quantum computing. Dr. Medalsy details the technical and strategic breakthroughs made possible by these digital twins, discusses the impact on quantum algorithm and hardware development, and offers insights into the future of the field.

Key Discussion Points & Insights

1. Dr. Izhar Medalsy's Background & Quantum Elements’ Mission

• Personal & Academic Journey ([01:03])

  • Medalsy studied physics and chemistry in Israel, focusing on nanoscale computation.
  • Postdoc at ETH Zurich; moved from research to business and product roles; founded startups.
  • Co-founded Quantum Elements with Prof. Daniel Lidar (quantum error correction, USC) and Prof. Amira Kobe (experimental quantum, Harvard).

• Genesis of Quantum Elements ([01:45])

  • Realized that for quantum computing to progress, classical computation must augment quantum hardware—creating "digital twins."
  • “You need to be able to use the ability of, to augment those system using classical devices, meaning creating a digital twins of those systems and best inbreed classical tools that are used today for software development in order to control them, accelerate their development and ultimately enable end user applications.” (A, [02:41])

2. What is a Digital Twin in Quantum Computing?

• Physics & Algorithmic Components ([06:01])

  • Build a mathematical and physical description of the quantum device (e.g., superconducting, ion traps).
  • Solve the quantum equations using large-scale classical computers to simulate real quantum hardware.

• Scalability Breakthroughs ([06:57])

  • Prior tools might model 1–2 qubits; their methods scale up to 50–100 qubits with realistic noise —a crucial advance.

• Environment and Noise Modeling ([09:02])

  • “What is the noise that is affecting the system and how does it evolve over time as you are operating this machine?” (A)
  • All relevant environmental factors—noise, crosstalk, leakage—are baked into the digital twin.

3. Hardware-Specific vs. General Purpose Digital Twins

• Level of Detail Depends on Use Case ([12:51])

  • When working closely with hardware vendors, twins are fine-tuned to match precise operating conditions.
  • For others (app developers, upper software stack), more abstracted twins based on available or output data suffice.

• Validation Example: IBM’s Shor’s Algorithm ([16:03], [00:25], [23:51])

  • IBM’s qubits + Shor’s algorithm (factoring 21) yielded ~80% accuracy.
  • Quantum Elements’ digital twin identified “crosstalk” bottleneck, engineered suppression, raised real-hardware accuracy to 99%.
  • Notable quote: “We took this knowledge that we gained from our digital twin and implemented it on the IBM platform and got to 99% accuracy on Shor's algorithm.” (A, [00:25], recapped [16:40], [23:51])

4. Modeling Errors and Noise: How & Why

• Core Noise Models & Physics ([21:17])

  • Models solve the quantum master equations directly.
  • Key error/noise parameters: T1, T2, T*, 1/f noise, crosstalk, leakage.
  • “Crosstalk is a very notorious one… When you have 100 [qubits], it becomes a bit harder. When you have a thousand, well, forget it.” (A, [21:48])

• Impact ([22:53])

  • Accurate physical modeling at scale enables not just simulation but informed, targeted optimization and error correction.

5. The Role & Power of AI

• Data as a Bottleneck ([26:58])

  • Training AI for quantum error correction needs massive, high-quality, labeled data—but quantum hardware is scarce and variable.
  • Digital twins generate that data, across current and hypothetical future hardware, swiftly and cheaply.

• AI-Powered Optimization ([26:58], [31:31])

  • Example: Error correction, syndrome extraction; digital twins let you label hardware configurations and error manifestations, then use AI to minimize them.

  “What you… The advantage that you have with the digital twin is that you're not bound by any physical hardware. You can not only create a representation of the current state, but also of future state to a system scale that is big enough that allows you to generate massive amount of data…” (A, [27:20])

  • AI closes the loop: proposes mitigations, re-tests, and continually adapts as hardware or conditions evolve.

6. User Pipeline: From Algorithm to Hardware Optimization

• User Workflow Example ([34:18])
  • User specifies hardware/platform, uploads a quantum circuit, sets noise/environmental parameters.
  • Runs simulations, injects remedies (e.g., error suppression pulses), iterates, and trains AI to propose optimizations.
  • Cost advantage: real quantum hardware is expensive, while digital twin experiments are far cheaper and scalable.

7. Demonstrated Outcomes & Quantum Industry Impact

• Before & After Improvements ([38:25])

  • Made IBM hardware capable of logical qubits—previously only shown by Google.
  • Pushed Rigetti hardware to 99.9% fidelity via optimized configuration.
  • General result: Enables noisy current hardware to do more meaningful, larger-scale computation.

• Role as a Quantum Benchmark ([41:36])

  • Digital twins can serve as third-party validators for claims of “quantum advantage.”
  • But, just as in classical hardware, simulation at some scale hits limits, so both hardware and simulation advance in tandem.

• Scaling Limits & Frontiers ([44:16], [46:46])

  • Currently, digital twins can simulate 100+ noisy qubits. Industry expects the gap to hover in the few hundred-qubit range for the foreseeable future.
  • Classical tools will continue to drive quantum hardware R&D: “It’s clear that classical devices are the tools to push quantum forward.” (A, [45:32])

8. Modalities & Transferability

• Current Modalities: ([47:32])

  • Focus on superconducting, trapped ion, and spin qubits; photonics and others are being explored.
  • As abstraction increases (higher up the stack), some AI/data tools become portable across hardware types.

• Primary Customers: ([49:03])

  • Hardware designers: optimize and calibrate next-gen devices.
  • Algorithm/circuit developers: reduce errors, improve circuit performance.
  • Error correction researchers.
  • Anticipates eventual direct end-user (e.g., pharma, finance) adoption in combination with AI.

9. IP, Competitive Dynamics, and Vendor Neutrality

• Balancing Openness & Proprietary Info ([52:03])

  • “We will give you the best engine to now go and run this twin at scale for your development… we kind of have this firewall between their secret sauce and their know-how and their IP and our capability. And I think it works very nicely.” (A, [52:53])

• Ecosystem Effects ([51:25])

  • As more vendors optimize via digital twins, performance gaps and choices become clearer—raising the bar for quantum hardware.

10. Roadmap & Future Milestones

• Optimizing Out of Demand? ([54:02])

  • Improvement is a moving target: as quantum scales, the need to squeeze more from hardware and algorithms will continue.

• Timeline to Practical Quantum ([55:00])

  • “Noise level of the system, number of qubits, and the number of qubits that are needed in order to run a specific algorithm. In all of those axises, we're seeing an acceleration beyond what we expected… I don't think there will be a step function… What I think we will see is that Quantum is adding benefit sooner through some hybrid approach…” (A, [55:09])

• AI & Quantum—Hype vs. Real Promise ([56:53])

  • Skeptical on near-term quantum computers running large-scale AI, but optimistic on AI as an accelerator for quantum.

• Milestones to Watch (Next 3–5 Years) ([58:42])

  • Noise/fidelity metrics—esp. two-qubit gate fidelity.
  • Physical and logical qubit counts (e.g., IBM aims for 10,000 physical qubits by 2030).
  • Ratio and conversion efficiency from physical to logical qubits.
  • Demonstrations of practical quantum advantage and sustained algorithmic progress.

Notable Quotes & Moments

• On quantum error correction and scaling digital twins:
  “If you're able now to solve 50 or 100 of those qubits, you can now tap into every problem that the industry is tackling…” (A, [07:00])

• On using AI with digital twins:
  “One of the things that AI needs… is large scale curated data labeled data that is actionable and can be used to solve a specific problem.” (A, [26:59])

• On the control that digital twins offer over quantum systems:
  “Because, you know, one, one of the things that we always say about the quantum system is the moment you ask the quantum system a question, it collapses… But you can do that with a simulator.” (A, [32:51])

Important Timestamps

• [01:03] – Izhar Medalsy: Academic and professional background
• [03:50] – What is a digital twin in quantum computing?
• [06:01] – Physics and algorithmic components of digital twins
• [09:02] – Simulating environment and noise in digital twins
• [12:51] – Levels of abstraction: hardware vs. application layer users
• [16:03], [23:51] – IBM Shor’s algorithm: from 80% to 99% accuracy
• [21:17] – Detailed noise modeling (T1, T2, crosstalk, leakage)
• [26:58] – AI’s role: data generation, training, and large-scale error correction
• [34:18] – Walkthrough of user workflow: from algorithm upload to hardware optimization
• [38:25] – Other “before and after” improvements beyond IBM Shor’s
• [41:36] – Digital twins as independent arbiters of quantum advantage
• [44:16] – Scaling limits and advances in digital twin simulation
• [47:32] – Modalities: superconducting, ions, spins, photonics
• [49:03] – Customer segments: hardware, algorithm, and end users
• [55:00] – Forecast: when quantum hardware becomes truly useful
• [58:42] – Key industry milestones for the next 3–5 years

Closing Thoughts

Dr. Medalsy provides a clear, jargon-light exploration of how digital twins and AI are turning quantum's fundamental frailty (noise, instability) into sources of rapid innovation. His optimism is grounded in a pragmatic, tool-building ethos: “It's not a question of whether or not [classical devices] will participate, it's clear that classical devices are the tools to push quantum forward.” (A, [45:32])

The outlook? More rapid progress than most expect, provided these deep digital-physical loops—connecting theory, simulation, hardware, and AI—continue to accelerate.

End of Content Summary