- Introduced an LLM-guided experimentation loop that searches adaptive solver-control policies for variational quantum optimization.
- Evaluated on MIS and CVRP, showing that staged confirmation is essential to avoid proxy-evaluation failures.
Building quantum algorithms and AI systems across optimization, simulation, scientific discovery, agents, and language-model workflows — from Shor's algorithm and quantum differential-equation solvers to hybrid quantum–classical software.
I am a Research Engineer at Singapore Management University working across quantum computing, quantum algorithms, and AI. My work spans hybrid quantum-classical optimization, Shor-style factoring experiments, quantum simulation and differential-equation solvers, LLM-guided research loops, AI agents, and scientific software systems.
I worked on MettleQ, a trustworthy local quantum-circuit simulation package for Apple Silicon. It integrates with Qiskit and PennyLane, exposes exact statevector and matrix-product-state simulation paths, and uses MLX/Metal execution with measured dispatch policies and reproducible benchmark evidence.

Ratio above 1.0x means MettleQ GPU was faster. In this latest run, MettleQ crosses PennyLane Lightning at 16 qubits and Qiskit Aer at 20 qubits.
| Qubits | Qiskit CPU/GPU | PennyLane CPU/GPU |
|---|---|---|
| 20 | 1.327x | 11.097x |
| 24 | 4.554x | 18.713x |
| 29 | 2.739x | 16.005x |
I also built The Atelier Lab, a local AI research laboratory for reproducing modern AI systems from first principles. It combines a foundation-model track with an autonomous local agent that runs on a single MacBook using Ollama, local embeddings, RAG, durable memory, and test-verified build mode.
Baseline training measurements on an Apple M3 Pro, alongside a local agent stack for RAG, memory, code repair, and test-verified edits.
| Model | Throughput | Run cost | Loss |
|---|---|---|---|
| 73.5M | ~18.7k tok/s | ~1.2 h | 1.1664 BPB |
| 286.2M | ~4.47k tok/s | ~5.1 h | 1.0954 BPB |
A non-technical walkthrough of maritime energy resilience, chokepoint disruptions, and stochastic optimization for national supply networks.
Read on Medium ↗ N° 02 · 2026India's 1,000 km quantum communication demonstration, what QKD changes, and why quantum security is becoming infrastructure.
Read on Medium ↗ N° 03 · 2026A look at ambitious new resource estimates for quantum factoring and what they imply for Shor's algorithm at cryptographic scale.
Read on Medium ↗ N° 04 · 2026What it would actually take to break RSA with a quantum computer, from hardware assumptions to fault-tolerant overheads.
Read on Medium ↗ N° 05 · 2026An honest-metrics tour through published Shor's algorithm results and the classical post-processing baselines behind them.
Read on Medium ↗ N° 06 · 2026A hands-on deep dive into running Shor's factoring algorithm on IBM quantum hardware and the limits visible today.
Read on Medium ↗ N° 07 · 2025Exploring the complexity of the MDKP and how quantum algorithms attempt to tackle this constrained optimization challenge.
Read on Medium ↗ N° 08 · 2025A deep dive into the MIS problem, its importance in network analysis, and the potential of quantum approaches.
Read on Medium ↗ N° 09 · 2025Analyzing the QAP, known for its extreme computational difficulty, and benchmarking quantum solvers against it.
Read on Medium ↗Open to research collaborations, talks, and hard problems across quantum computing, algorithms, agents, and scientific AI.