QFlowNet: Fast, Diverse, and Efficient Unitary Synthesis with Generative Flow Networks
Inhoe Koo, Hyunho Cha, and Jungwoo Lee
IEEE International Conference on Quantum Communications, Networking, and Computing (QCNC), 2026.
Inhoe KooQuantum computing · AI · SNU
Quantum curiosity.
Intelligent possibilities.
I'm Inhoe, exploring quantum computing, AI, and the physics they can reveal.
Every idea is connected. Choose an orbit to discover mine.
About meProfile & interests
PublicationsPapers & talks
Vision & goalsPurpose & direction
EducationAcademic journey
ResearchIdeas & experiments
Beyond the labLife & community
A JOURNEY THROUGHSeoul National UniversityETH ZürichPaul Scherrer InstituteIonQ
THE BIGGER QUESTION
What can quantum machines
help us understand?
My goal is to develop quantum computational methods for physical simulations beyond classical reach—bringing algorithms, learning, and hardware into the same conversation.
Discover my vision & goals ↗IDEAS INTO IMPACT
A closer look at my work.
QUANTUM CIRCUITS01 /
QFlowNet
Finding a diverse landscape of compact quantum circuits with generative flow networks.
IEEE QCNC 2026
QUANTUM × RL02 /
CASH: Learning to measure
Making every measurement count through adaptive, intelligent quantum-state acquisition.
Submitted · IEEE QCNC 2027
NEUTRAL ATOMS03 /
RL for Atom Array
Exploring reinforcement learning for efficient rearrangement of neutral atom arrays.
JSAP oral presentation · 2026
01 / ABOUT ME
A curious mind.
A quantum perspective.
Inhoe Koo / Inae
I'm an Electrical and Computer Engineering undergraduate at Seoul National University, exploring quantum computing, artificial intelligence, and quantum simulation.
My work connects learning-based algorithms with physical quantum systems—from generative circuit synthesis and adaptive measurements to neutral atoms and superconducting hardware.
My toolkit
PythonC / C++QiskitPyTorchTensorFlowReinforcement learningGFlowNets
Let's connect
I'm always interested in thoughtful conversations about quantum computing and AI.
02 / PUBLICATIONS & PRESENTATIONS
Research, shared.
Papers, ongoing work, and conversations across the quantum community.
CASH — adaptive quantum state classification
Reinforcement learning for selecting Pauli measurements from cumulative outcomes. Submitted to IEEE QCNC 2027.
Explore the project ↗Reinforcement Learning Enhanced Algorithm for Fast Atom-Array Generation
Inhoe Koo, Luis Fernández, and Wenchao Xu
Japan Society of Applied Physics (JSAP), 2026.
Unitary Synthesis using GFlowNets
Inhoe Koo
IEEE/IEIE International Conference on Consumer Electronics Asia (ICCE-Asia), 2025.
Quantum AI trading project
SNU KIC SV K-BioX ABDD SUMMIT, Global Innovation Symposium, Stanford University.
03 / VISION & GOALS
Understand nature.
Expand what is possible.
My goal is to develop quantum computational methods for physical simulation problems beyond the reach of classical computers.
MY RESEARCH PHILOSOPHY
Useful quantum computation emerges when algorithms, hardware constraints, and physical implementation are designed together. I use AI to navigate complex design spaces, guided by physical understanding.
01
Build learning-based quantum compilers
I want to connect quantum simulation with executable circuits: learning how to choose gate sequences, qubit mappings, and schedules that balance fidelity, circuit depth, and compilation time.
My starting point: QFlowNet ↗02
Design around the realities of quantum hardware
I am interested in compilers for three-dimensional atom arrays and fault-tolerant architectures that account for restricted movement and stochastic atom loss. Hardware constraints should shape the algorithm from the beginning.
Learning from neutral-atom systems ↗03
Connect control, measurement, and physical insight
From adaptive measurement policies to superconducting RF components, I aim to improve how quantum information is delivered, protected, and read out. My interests include backend-aware compilation that balances depth, ancillas, connectivity, and synchronization.
Explore my research journey ↗04 / EDUCATION
Foundations for
what comes next.
MAR 2021 — PRESENT
Seoul National University
B.S. in Electrical and Computer Engineering
GPA 4.17 / 4.3Major GPA 4.16 / 4.3
Includes 21 months of R.O.K. mandatory military service, May 2023 – February 2025.
FEB 2026 — AUG 2026
ETH Zürich
Exchange student
Swiss Federal Institute of Technology Zürich
MAR 2018 — FEB 2021
Seoul Science High School
GPA 4.24 / 4.3
05 / RESEARCH
Where quantum
meets learning.
From circuit discovery and atom movement to superconducting control and readout.
SEP 2026 — PRESENT
Applied Superconductivity Lab
SNU · Research intern · Prof. Seungyong Han
Electromagnetic and thermal simulations of superconducting RF components for qubit control and readout. I study resonators, transmission lines, and Purcell filters to understand quality factors, coherence, and readout fidelity under cryogenic operation.
FEB 2026 — AUG 2026
Experimental Quantum Engineering Lab
ETH Zürich · Semester project · Prof. Wenchao Xu
Hardware-aware reinforcement learning for rearranging neutral Rubidium and Ytterbium atoms with acousto-optic deflectors. Research conducted at the Paul Scherrer Institute, Switzerland, and presented orally at JSAP 2026.
Explore the project ↗JAN 2025 — JAN 2026
Cognitive Machine Learning Lab
SNU · Research intern · Prof. Jungwoo Lee
QFlowNet for generative unitary synthesis, and CASH for adaptive Pauli measurement selection. QFlowNet was presented at IEEE QCNC 2026; CASH has been submitted to IEEE QCNC 2027.
FEB 2025 — APR 2025
IonQ Spring Research Program
Researcher · Adviser: Guillermo Aboumrad
Proposed a curriculum-learning model for discovering an ansatz to estimate the lowest energy of a Hamiltonian.
06 / BEYOND THE LAB
More than
one dimension.
Teaching, community, and the interests that keep me curious.
Teaching & leadership
- Teaching assistant · SNUIntroduction to Electromagnetism, Sep–Dec 2025.
- Laboratory teaching assistant · SNUIntroduction to Circuit Theory, Mar–Jun 2025.
- STEM · Engineering Honor SocietyExternal Affairs, Sep 2025 – Dec 2026. Academic seminars and alumni homecoming events.
- AI Biohealthcare Drug Discovery SUMMITProtocol Team, Stanford University, Jan 2026.
Awards & honors
- Presidential Science ScholarshipKorea Student Aid Foundation (KOSAF), 2025
- 1st Prize · AI Product Recognition CompetitionSNU, 2022
- 2nd Prize · AI Resource Recycling ChallengeKorea University, 2022
- 3rd Prize · Big Data Innovation ContestSNU, 2021
- Academic Excellence ScholarshipSNU, 2021–2023
- Korean Physics Olympiad Winter SchoolKorean Physical Society, 2019
Outside the lab
PhilosophyQuantitative tradingAI financial modeling
Korean (native) · English (fluent, TOEFL iBT 109)
RESEARCH NOTE / QUANTUM CIRCUITS
QFlowNet.
Fast, Diverse, and Efficient Unitary Synthesis with Generative Flow Networks
IEEE QCNC 2026GFlowNetsTransformers
The question
How can we discover many compact quantum circuits for a target unitary, instead of converging on a single solution?
The approach
QFlowNet pairs generative flow networks with Transformers to sample circuit candidates in proportion to their reward. Synthesis becomes a path-finding problem: reduce the unitary residual to the identity matrix.
U V† → … → I
Why it matters
The framework combines circuit diversity, compactness, and fast sampling. The project's research note reports a 99.7% success rate on 3-qubit benchmarks with circuit lengths of 1–12.
Inhoe Koo, Hyunho Cha, and Jungwoo Lee · IEEE QCNC 2026
CASH / ADAPTIVE QUANTUM STATE CLASSIFICATION
Learning to
measure.
Making each measurement more informative through reinforcement learning.
Submitted to IEEE QCNC 2027Adaptive measurements
The question
Fixed measurement bases can waste the limited shots available from an experiment. Can a learning agent decide which information to acquire next?
The approach
I developed CASH with colleagues at SNU's Cognitive Machine Learning Lab. The framework sequentially selects Pauli measurement operators from cumulative outcomes to maximize information gained per shot.
What we studied
We evaluated adaptive state classification across GHZ, W, cluster, and noisy parameterized states, comparing against static-basis measurements under the same measurement budget.
The direction
This work connects learning-based decision-making with experimental information acquisition. The manuscript has been submitted to IEEE QCNC 2027.
All research ↗SEMESTER PROJECT / ETH ZÜRICH & PSI
RL for
Atom Array.
Reinforcement learning for neutral-atom rearrangement.
From learning to the laboratory
At the Experimental Quantum Engineering Lab, ETH Zürich, I developed reinforcement learning algorithms for the efficient rearrangement of neutral Rubidium and Ytterbium atoms using acousto-optic deflectors.
The semester project was advised by Prof. Wenchao Xu, with research conducted at the Paul Scherrer Institute in Switzerland. I presented this work orally at JSAP 2026.
Let the hardware shape the algorithm
AOD systems impose row/column motion and non-crossing constraints. I developed learning-based packing actions compatible with those constraints, connecting path planning with physically executable atom movement.
February – August 2026Neutral atomsReinforcement learning