This is,

Xiaoxuan Zhang.

Software engineer with years of experience in C++, Golang, blockchain infrastructure, and crypto trading systems. Currently focused on applying AI and transformer-based time series models to cryptocurrency markets, bridging quantitative research with real-world trading infrastructure.

Resume

Experience

TSFM Researcher (Master's Thesis) - University of Amsterdam
2024 - Present
  • Researching Time Series Foundation Models (TSFMs) with a focus on the Chronos model family — pre-trained probabilistic forecasters built on language model architectures — as the core subject of the master’s thesis at UvA.
  • Applying Chronos2 to cryptocurrency price series for multi-horizon trend forecasting, evaluating zero-shot and fine-tuned performance on volatile, non-stationary financial data.
  • In-depth understanding of probabilistic forecast objectives, especially pinball (quantile) loss, with hands-on analysis of their behavior on fat-tailed distributions.
  • Investigating the transferability and fine-tuning generalization of large-scale TSFMs for forecasting cryptocurrencies such as BTC, ETH, and LTC under rapidly shifting market regimes.
Senior Software Engineer - Mavenir
Feb 2021 - June 2024
  • Developed Golang-based telecom network element management microservices for virtual RAN components such as RU and DU, covering onboarding, version validation, image pull/update workflows, traffic forwarding, and fast outage recovery.
  • Built and operated a Kubernetes/Docker-deployed microservices cluster using NATS for internal service messaging and Kafka for inter-cluster communication, traffic smoothing, and producer-consumer workload decoupling.
  • Led high-availability upgrades by compressing and externalizing legacy service state into Couchbase, enabling stateless microservices and improving recovery across multi-instance deployments.
  • Designed a multi-instance resilience model with shared-memory state synchronization and a load-balancer gateway, strengthening disaster prevention, failover behavior, and service availability under telecom-grade reliability requirements.
  • Developed a Golang concurrency middleware to control unbounded goroutine growth, combining a channel-based worker pool with dynamic worker scaling and closure-based callback registration for asynchronous, non-blocking event processing.
Independent Quantitative Trading Developer - GeekChomolungma
2019 - Present
  • Built a real-time market data pipeline syncing order book snapshots, trades, and OHLCV streams from Binance into MongoDB (cold storage) and Redis (hot cache), forming the data infrastructure layer for all downstream strategy research.
  • Ported and adapted statistical trading strategies from TradingView Pine Script — covering momentum, volatility breakout, and candlestick pattern signals — into a live execution framework on Binance, with real-money validation.
  • Fine-tuned quantile-regression-based time series foundation models on cryptocurrency price series to capture asymmetric return distributions; constructed and backtested AI-driven alpha signals derived from predicted quantile spreads and directional probabilities.
  • Completed Optiver’s market-making training program, gaining structured exposure to delta-neutral positioning, and bid-ask spread management.
Core Developer - Lava
Nov 2018 - Dec 2020
  • Led the full lifecycle development of Lava, a proof-of-capacity (PoC) based cryptocurrency forked from Bitcoin Core.
  • Architected and implemented consensus modifications to support PoC mining, enhancing energy efficiency compared to PoW.
  • Oversaw system integration, wallet and miner tools, and network bootstrapping.

Projects

AI for Financial Time Series

Chronos2 Crypto Forecasting
Chronos2 TSFM Quantile Forecasting
Chronos2 Crypto Forecasting
Fine-tuning and applying Chronos2, a time-series foundation model, to cryptocurrency price forecasting. The project explores probabilistic multi-horizon prediction on volatile crypto data, with the BTC chart shown here as an illustrative output that changes with training mode, data window, and evaluation setup.
Risk-Calibrated Crypto Signal Framework
Custom Loss Hawkes Process Backtesting
Risk-Calibrated Crypto Signal Framework
Master's thesis project that extends the forecasting pipeline into a trading-signal and backtesting framework. It combines custom loss design for quantile prediction with Hawkes-process market risk modeling, then uses risk-adjusted expected return as a timing signal. The full thesis is linked in the Articles section below.

Distributed Systems

Telecom Network Element Management System
Golang Kubernetes NATS/Kafka Telecom
Telecom Network Element Management System
Proprietary Mavenir microservices platform for managing virtual telecom network elements such as RU and DU across onboarding, version validation, image pull/update workflows, traffic forwarding, and fast outage recovery. Built in Golang and deployed on Kubernetes/Docker, the system used NATS for internal messaging and Kafka for inter-cluster communication, traffic smoothing, and producer-consumer decoupling.
Gofer
Golang Concurrency Worker Pool Async Callbacks
Gofer
Proprietary Golang concurrency middleware developed at Mavenir for high-concurrency telecom microservices. Gofer uses a two-layer design: a channel-based worker pool with dynamic worker scaling at the bottom, and a closure-based asynchronous callback framework at the top. It prevents uncontrolled goroutine growth while keeping business workflows event-driven, elastic, and non-blocking.

Crypto Trading Infrastructure

ChomoSyncer
C++ Redis/MongoDB Binance API
ChomoSyncer
A high-performance Binance market data synchronizer that streams multi-symbol cryptocurrency Kline data in real time, distributes updates through Redis Streams, and persists historical data into MongoDB for research, backtesting, and downstream trading systems.
Binance Orderbook Visualizer
Python Binance WebSocket
Binance Orderbook Visualizer
A real-time 3D visualization app for Binance order books, using streaming depth data to reveal market microstructure, liquidity shifts, best bid/ask trajectories, and transient order-book imbalances.
Kline Preprocess
OHLCV TA
Kline Preprocess
A cryptocurrency K-line preprocessing pipeline that standardizes Binance OHLCV data and generates no-lookahead technical indicator features for downstream consumers such as traditional price-volume models and time-series AI foundation models.

Market Making & Derivatives

Optiver Academy Market Making
Market Making Optibook Risk Engine
Optiver Academy Market Making
A one-month Optiver Academy market-making project refactored into a reusable Optibook strategy framework. It implements passive cross-instrument quoting across ETFs, spot products, futures, and index-linked instruments, with shared pricing, quote planning, execution, and risk engines organized around an exchange-state-plan-execute workflow.
Option Pricing
Option Pricing Volatility Smile
Option Pricing
Numerical methods for option pricing — volatility smile modeling and computational finance assignments.

Blockchain

Lava
C++ Bitcoin
Lava
A proof-of-capacity (PoC) based cryptocurrency forked from Bitcoin Core.

Articles

[1] Twintinuum: Advancing Self-Calibrating Physical-Digital Continuum
Zhiheng Yang, Xiaoxuan Zhang, Adam Belloum, Chrysa Papagianni, Paola Grosso
IEEE INFOCOM 2026 · 2026 — Top-tier networking & communications conference (CORE A*)
[2] The Simple Math Behind Foundation Models(e.g. LLM): Cross-Entropy, Not Magic
Xiaoxuan Zhang
LinkedIn Pulse · Sep 9, 2025
[3] Forecasting, Uncertainty and Excitation: A Unified Framework for Cryptocurrency Trading
Xiaoxuan Zhang
Master's Thesis, University of Amsterdam · 2026

Education

2024 - 2026
Master of Computational Science
University of Amsterdam
GPA: 7.7/10.0
  • Built a solid foundation in stochastic differential equations (SDEs), Monte Carlo simulations, and numerical methods for option pricing.
  • Gained in-depth understanding of numerical PDE solvers, discretization techniques, and iterative schemes in the context of scientific computation.
2013 - 2015
Master of Control Engineering
Wuhan University
GPA: 3.5/4.0
  • Core courses included Matrix Theory, Control Theory, Numerical Analysis, and Optimization.
  • Proficient in modeling and simulation of industry processing, focusing on industry efficiency solutions by regression and classification tools like SVM, LS-SVM, and PSO.
2009 - 2013
Bachelor of Automation
Wuhan University
GPA: 3.38/4.0

Feel free to reach out.

Mail - qqww574817827@gmail.com