Fu-Te Wong [contact removed] · [contact removed] · Toronto, Canada · github.co · HuMetric

Experience

Researcher — Vector Institute for Artificial Intelligence

2024 → Present

Built a multi-agent LLM/RL pipeline for automated quantum machine-learning algorithm search; abstracted a novel advanced for-loop construct accepted at QTML 2025 (arXiv:2506.18260). Designed an agentic search framework with modular tool use, prompt chaining, and self-evaluation loops; architected the inference system; and authored a quantum backpropagation formulation, under review. Formulating artificial consciousness via Integrated Information Theory. Co-organiser of an AI-for-Reasoning reading group; driving LLM multi-token prediction on multi-hop reasoning and suicidality-detection agent projects; developing a DPO variant.

Ph.D. Researcher — CS & Medical Science — University of Toronto

2021 → 2026

PhD thesis: Agentic system for exploration of algorithmic ANNs. Contributed to the generative quantum eigensolver (GQE) (arXiv:2401.09253). Engineered virtual brain stimulation AI; integrated E-field modelling with connectome tractography for TMS target selection.

Teaching Assistant — Deep Learning — Neuromatch Academy

2020 → 2023

Lead TA in 2021 for a global cohort of 1,000+ students. Delivered a deep-learning curriculum using PyTorch and led debugging sessions and model evaluation workshops across distributed online cohorts.

Data Analyst — Institute of Linguistics, Academia Sinica

2019 → 2021

Designed and deployed multimodal neuroimaging pipelines covering MRI, fMRI, DTI, and MEG for a cross-age working-memory study. Reduced manual quality-control time by approximately 40% through automated topological analysis scripts.

Research Assistant — Medical AI — Center for AI in Medicine, Taipei Medical University

2018 → 2019

Delivered a CNN-based diagnostic model for idiopathic normal-pressure hydrocephalus and brain-tumour segmentation. Published a CNN suicidality-prediction model in the Journal of Clinical Psychiatry, 82(2), 2021. Built and documented an end-to-end data pipeline from DICOM ingestion through preprocessing and model inference to clinical reporting.

Software Engineer — Cybersecurity Technology Institute, Institute for Information Industry

2017 → 2017

Developed security tooling and data-processing scripts in Python and C++ within an agile team. Participated in code review and issue-tracking workflows.

Visiting Student — Big Data & ML Infrastructure — Modeling & Informatics Lab, National Taiwan University

2016 → 2017

Provisioned and administered a Hadoop/Spark CDH cluster for large-scale neural image analysis using the Thunder project and TensorFlowOnSpark. Aligned chatbot LU/NLG embeddings with brain activity signals.

Clinical Psychologist — Taipei Veterans General Hospital

2011 → 2013

Worked in the Psychiatry Department, Neurological Institute, and Rehabilitation Center.

Featured projects

MetaScientist (Meta_Scientist)

Python · MetaGPT · Monte Carlo Tree Search · Docker

Self-improving agentic research framework pairing LLM multi-agent orchestration with Monte Carlo tree search (MCTS) to automatically search quantum machine-learning algorithm pipelines. Integrates web search, task planning, and a dual-loop research cycle. Implementation framework behind the QTML 2025 paper; extends MetaGPT/SELA.

DDB Knowledge Agent (ddb-knowledge-agent-py)

Python · FastAPI · Server-Sent Events · Gemini · PostgreSQL · pgvector · Apptainer

Streaming RAG agent fusing pgvector similarity retrieval with PostgreSQL full-text search via reciprocal rank fusion (RRF). Includes ReAct tool calling, query-complexity-tiered model routing, semantic caching, and cost auditing. PostgreSQL runs under rootless Apptainer and is deployable on Docker-less HPC clusters.

GOAT Network Agent (goat-agent_demo)

Node.js · Express · ERC-8004 · x402

Autonomous Web3 agent registering an ERC-8004 on-chain identity and integrating with the x402 payment flow. Executes price tracking, conditional triggering, on-chain payment, and attested settlement end-to-end with no human intervention after launch.

optlab — Derivatives Pricing & Strategy Backtesting (optlab-py)

Python · NumPy · SciPy · Pandas · pytest · GitHub Actions · Black-Scholes · CRR Trees · Monte Carlo Simulation · Heston Model · Andersen QE · Raw-SVI

Pricing library and vectorised backtester built to test whether the variance risk premium survives transaction costs. Includes Black-Scholes analytic Greeks, CRR trees with American exercise, variance-reduced Monte Carlo with a 12.7× improvement via antithetic and control variates, Heston characteristic-function pricing and Andersen QE simulation, implied-volatility inversion, and raw-SVI and Heston surface calibration. Across 24 simulated 10-year paths, delta hedging lifted short-straddle Sharpe from 0.24 to 0.49, while frictions consumed 70% of gross P&L institutionally and all of it at retail. Included 102 model-validation tests covering convergence orders, Greeks versus finite differences, no-arbitrage bounds, and P&L versus a hand-computed ledger.

Quantum Backpropagation

Quantum Machine Learning

Publication by Wong, F.-T. (2026). Under review.

Multi-Token Prediction and Alternative Tokenization Improve Multi-Hop Reasoning and Graph Search

LLMs · Multi-Token Prediction · Multi-Hop Reasoning · Graph Search

Publication by Frydenlund, A., Wong, F.-T., et al. (2026). Under review.

Advanced For-Loop for QML Algorithm Search

Quantum Machine Learning · Algorithm Search

Publication by Wong, F.-T. (2025), accepted at QTML 2025.

Generative Quantum Eigensolver (GQE)

Quantum Computing · Quantum Machine Learning

Publication by Nakaji, K., Wong, F.-T., et al. (2024).

CNN Suicidality Prediction

Convolutional Neural Networks · Medical AI

Publication by Chen, V. C.-H., Wong, F.-T., et al. (2021), Journal of Clinical Psychiatry 82(2).