Alston Hui · Resume
LinkedIn

Alston Hui

Jialong.hui@alumni.utoronto.ca · +86 178 0290 6132 · github.com/alstonxi · Toronto / Remote · English & Mandarin
Full-Stack AI / LLM Serverless Edge Linux Infra Quant Finance
University of Toronto HBSc · Statistics
Economics · Computer Science
QS Top 50 Global
Profile
Full-stack developer and AI systems engineer with a quantitative background (Statistics, Economics, CS — University of Toronto). Specialises in LLM integration, serverless edge architecture, and production AI pipelines. Combines deep financial market domain knowledge with hands-on infrastructure engineering to build systems that operate at the intersection of intelligence and commerce. Experienced in Linux server administration, encrypted network infrastructure (Shadowsocks, VLESS/Reality, 3X-UI), and cloud platforms (GCP, AWS). Proven ability to own complete systems from architecture design through deployment and live operation.
Projects
Algoter AI Messenger Sales Bot · Production System
AI-powered customer acquisition and sales automation — Facebook Messenger
2025 – Present
  • Serverless Edge Architecture: Designed a Cloudflare Workers webhook pipeline processing Messenger events in real time; sub-100ms P99 on FAQ paths, <8s end-to-end on AI inference.
  • 5-Phase Routing Engine: Built echo filtering → escalation detection → KV-backed FAQ matching with negation guard → Workers AI inference → async success-signal learning; ~80% FAQ resolution rate with zero false positives.
  • LLM Prompt Engineering: Engineered a 2,000-token system prompt embedding an 8-stage sales qualification framework, 9-language detection, correction-handling logic, and dynamic few-shot injection for llama-3.2-3b-instruct.
  • Adaptive Learning Pipeline: Async background process detects conversion signals (registered / deposited), compresses successful conversation patterns via LLM summarisation, and injects rotating success examples into live context for continuous self-improvement.
  • Dual Storage Architecture: Cloudflare KV for low-latency state management (conversation history, user profiles, 7-day TTL) + D1 SQLite for structured analytics — FAQ coverage gaps, conversion funnels, escalation rates.
  • Admin Tooling: Full !command interface for live KB updates, FAQ approval workflow, stats dashboards, and per-user session resets — zero-redeployment live operation.
Cloudflare Workers KV D1 Workers AI (llama-3.2-3b) Facebook Graph API Webhook JavaScript Wrangler CLI
Encrypted Proxy & Private Network Infrastructure
Multi-node encrypted proxy across GCP and AWS
2024 – Present
  • Server Provisioning: Deployed and managed Linux VPS instances (Ubuntu/Debian) across GCP and AWS; handled instance sizing, region selection, disk management, and automated startup scripts via systemd.
  • Encrypted Tunnel Protocols: Configured Shadowsocks (AEAD encryption), VLESS with XTLS-Reality (anti-detection TLS fingerprint spoofing), and VMess — each tuned for different threat models and performance requirements.
  • 3X-UI Panel Administration: Deployed and managed 3X-UI web panel for multi-user inbound management, traffic monitoring, subscription link generation, and certificate management via Let's Encrypt / ACME.
  • Network Security Hardening: Configured iptables / ufw firewall rules, disabled root SSH login, enforced key-based authentication, set up fail2ban for brute-force protection, and applied port-knocking for additional obscurity.
  • Performance & Reliability: Tuned BBR TCP congestion control, optimised kernel network parameters (sysctl), configured process supervision (PM2 / systemd), and implemented log rotation and uptime monitoring.
  • DNS & TLS Management: Managed custom domain routing with Cloudflare DNS, configured SNI-based routing, and automated SSL certificate renewal.
GCP AWS Linux (Ubuntu/Debian) Shadowsocks VLESS/XTLS-Reality VMess 3X-UI iptables fail2ban BBR Cloudflare DNS Let's Encrypt
Predictive Analytics & Dashboard Engineering
Statistical modelling and interactive BI systems
2023 – 2024
  • Built linear / logistic regression models in R and SAS forecasting sales performance and customer behaviour across multi-variable datasets (~85% accuracy).
  • Engineered interactive BI dashboards in Power BI and Tableau; built a real-time Python / Streamlit web application translating model outputs into actionable business decisions.
R SAS Python Streamlit Power BI Tableau SQL
Work Experience
Operations Manager — Financial Technology
Hangzhou WeiTu Data Technology Co., Ltd.
Nov 2025 – Present
  • Client Infrastructure Management: Oversaw technical and operational client lifecycle across ECN/STP and Market Maker brokerage models for retail and HNW traders; implemented structured onboarding workflows, account activation protocols, and systematic retention processes.
  • CRM System Ownership: Owned and continuously optimised the CRM pipeline — built behavioural segmentation models, automated follow-up sequences, and conversion funnel analytics that directly drove measurable revenue improvement.
  • Data-Driven Market Communication: Designed and executed a content operations system for macro market analysis (NFP, CPI, FOMC, rate decisions) distributed across social channels, resulting in a 17% increase in community engagement.
  • Regulatory Compliance Operations: Managed end-to-end KYC/AML compliance processes under NFA/FCA/ASIC regulatory standards; maintained zero compliance incidents across all managed accounts.
  • Institutional Liaison: Served as the technical and commercial bridge between clients and LPs, Prime Brokers, and Banks — translating market microstructure and execution logic into client-facing communication.
  • Process Automation: Deployed the Algoter AI Messenger Bot in production to automate customer qualification and sales conversations — reducing manual response time by >90% while maintaining conversation quality.
Operations Manager
Carnex.ca — Tesla-Certified Dealership, Toronto
Dec 2024 – Jun 2025
  • Full-Cycle Operations Management: Owned end-to-end vehicle operations across procurement, inspection, arbitration, trade-in, and retail — managing cross-functional coordination between wholesale, retail, service, and finance teams.
  • System & Workflow Engineering: Redesigned core operational workflows using structured SOPs and process mapping; changes resulted in measurable cycle-time reduction and margin improvement within six months.
  • Data Analytics & Inventory Optimisation: Analysed CRM-based sales velocity, inventory ageing, and pricing data to drive dynamic pricing decisions and stock allocation — improving overall inventory turnover rate.
  • Regulatory & Compliance Execution: Liaised directly with Toronto government agencies on vehicle registration, safety certification, and title transfer — ensuring full regulatory compliance with Ontario MTO and OMVIC standards.
  • Team Leadership & Training: Recruited, onboarded, and supervised junior operations staff; implemented structured training programmes and performance evaluation frameworks.
  • Vendor & Client Relationship Management: Maintained and expanded wholesale procurement relationships; managed high-value client interactions with structured CRM follow-ups to drive referrals and repeat business.
AI Model Evaluator (Remote)
Outlier.ai
Oct 2024 – Jun 2025
  • Designed and iterated evaluation frameworks for large language models across instruction-following, factual accuracy, logical consistency, and safety dimensions using complex, multi-scenario prompts.
  • Conducted systematic root-cause analysis on model failure modes and delivered structured, data-driven feedback to model optimisation and prompt engineering teams.
  • Built reusable evaluation methodology documentation covering edge case identification, adversarial prompting patterns, and scoring rubric design.
Technical Skills
AI / LLM
Prompt engineering, few-shot learning, Workers AI, LLM evaluation, Anthropic Claude API
Cloud Infrastructure
Cloudflare Workers / KV / D1 / Pages, GCP, AWS, Linux (Ubuntu/Debian), SSH, systemd
Network & Security
Shadowsocks, VLESS/XTLS-Reality, VMess, 3X-UI, iptables, ufw, fail2ban, BBR, Let's Encrypt
Languages
JavaScript (Node.js), Python, R, SQL, Bash, SAS
APIs & Integrations
Facebook Graph API, Messenger Webhook, REST, Cloudflare Workers AI, Wrangler CLI
Analytics
Power BI, Tableau, Streamlit, regression modelling, CRM analytics

惠嘉龙

Jialong.hui@alumni.utoronto.ca · +86 178 0290 6132 · 多伦多 / 远程 · 英语 / 普通话
全栈开发 AI / LLM 边缘架构 Linux 运维 量化金融
多伦多大学 理学荣誉学士
统计学 · 经济学 · 计算机科学
QS 全球前 50
个人概述
全栈开发工程师，专注 AI 系统集成与 LLM 工程化落地。多伦多大学统计学、经济学与计算机科学背景，具备量化金融与金融市场系统性认知。擅长 Linux 服务器运维、加密网络基础设施搭建（Shadowsocks / VLESS-Reality / 3X-UI）及 Cloudflare 边缘架构设计，具备从架构设计到生产部署的完整工程链路经验。在 Outlier.ai 参与 LLM 评估工程，积累了系统性模型质量分析与 Prompt 优化方法论，并将 AI 能力应用于实际商业系统的生产环境部署。
项目经历
Algoter AI Messenger 销售自动化系统 · 生产环境
面向黄金量化交易公司部署的 AI 驱动客户获取系统，运行于 Facebook Messenger
2025 – 至今
  • 边缘架构设计：基于 Cloudflare Workers 构建 Serverless Webhook 处理管道，FAQ 路径 P99 延迟 <100ms，AI 推理端到端 <8s。
  • 五阶段路由引擎：实现消息过滤 → 升级检测 → KV FAQ 匹配（含否定词过滤）→ Workers AI 推理 → 异步成功信号学习，FAQ 命中率约 80%，零误触发。
  • LLM Prompt 工程：设计 2,000 token 系统提示词，嵌入 8 阶段销售资质框架、9 语言检测、纠正处理逻辑及动态 few-shot 成功案例注入。
  • 自适应学习管道：异步后台任务检测转化信号（注册 / 存款），通过 LLM 压缩成功对话模式，实时注入后续对话 context 持续优化。
  • 双层存储架构：Cloudflare KV（低延迟状态管理，7 天 TTL）+ D1 SQLite（结构化分析：FAQ 覆盖缺口、转化漏斗、升级率）。
  • 管理工具链：完整 !command 指令界面，支持知识库实时更新、FAQ 审批流、数据统计看板及用户会话重置，零重部署在线运营。
Cloudflare Workers KV D1 Workers AI (llama-3.2-3b) Facebook Graph API JavaScript Wrangler
加密代理与私有网络基础设施
跨 GCP / AWS 多节点加密代理网络
2024 – 至今
  • 服务器运维：在 GCP / AWS 上部署管理 Linux VPS（Ubuntu/Debian），涵盖实例配置、磁盘管理、区域选型及 systemd 自动化启动脚本。
  • 加密隧道协议：配置并维护 Shadowsocks（AEAD 加密）、VLESS + XTLS-Reality（TLS 指纹伪造抗检测）、VMess，针对不同威胁模型分别调优。
  • 3X-UI 面板管理：部署并管理 3X-UI Web 面板，实现多用户 Inbound 管理、流量监控、订阅链接生成及 Let's Encrypt / ACME 证书自动续签。
  • 网络安全加固：配置 iptables / ufw 防火墙规则，禁用 root SSH 登录，强制密钥认证，部署 fail2ban 防爆破，实施端口敲门增强隐蔽性。
  • 性能调优：启用 BBR TCP 拥塞控制，调整内核网络参数（sysctl），配置 PM2 / systemd 进程守护，部署日志轮转与在线监控。
GCP AWS Linux Shadowsocks VLESS/XTLS-Reality 3X-UI iptables fail2ban BBR Cloudflare DNS Let's Encrypt
数据分析与预测建模
统计模型与交互式 BI 系统
2023 – 2024
  • 使用 R、SAS 构建线性 / 逻辑回归模型，预测销售表现与客户行为（准确率约 85%）。
  • 使用 Power BI、Tableau 构建交互式分析仪表板；基于 Python + Streamlit 构建实时 Web 应用，将模型输出转化为可操作的业务决策支持。
R SAS Python Streamlit Power BI Tableau SQL
工作经历
运营经理 — 金融科技
杭州维图数据科技有限公司
2025年11月 – 至今
  • 客户系统管理：主导跨 ECN/STP 与 Market Maker 经纪模式的客户全生命周期管理，设计结构化入职流程、账户激活协议与系统化留存机制，覆盖零售及高净值群体。
  • CRM 系统建设：主导 CRM 管道的持续迭代优化，构建客户行为分层模型、自动化跟进序列与转化漏斗分析体系，直接驱动业务转化与收入增长。
  • 数据驱动内容运营：设计并执行宏观事件内容运营体系（非农、CPI、FOMC、利率决议），多渠道分发市场分析，社区互动率提升 17%。
  • 合规运营：在 NFA / FCA / ASIC 监管框架下，与合规及审计团队协同管理 KYC / AML 全流程，实现零合规事故记录。
  • 机构联络：作为客户与 LP、主经纪商、大型银行之间的技术与商务桥梁，将市场微观结构与流动性逻辑转化为客户沟通语言和内部决策依据。
  • AI 自动化部署：将自研 Algoter AI Messenger Bot 投入生产环境，实现客户资质筛选与销售对话自动化，手动响应时间压缩 >90%。
运营经理
Carnex.ca — 多伦多特斯拉认证经销商
2024年12月 – 2025年6月
  • 全链路运营管理：统筹车辆采购、评估、流转、置换与零售全链路运营，跨批发、零售、服务与金融多职能团队协调，在高节奏环境下平衡效率、风险与利润结构。
  • 流程重构与 SOP 建设：从零重构核心运营工作流，引入结构化 SOP 与流程映射方法论，六个月内实现可量化的周转周期缩短与利润率提升。
  • 数据分析与库存优化：基于 CRM 销售速度、库存老化与定价数据驱动动态定价决策与促销策略，提升整体库存周转率与运营效益。
  • 合规与政府对接：高频对接多伦多政府部门，推进车辆过户、安全认证与产权转让流程，确保全业务链符合安省 MTO 与 OMVIC 监管标准。
  • 团队培养与管理：主导初级运营人员的招募、培训与绩效评估体系建设，推行结构化培训计划，提升团队执行质量与整体产出效率。
  • 供应商与客户关系：维护并拓展多方批发采购关系；通过结构化 CRM 跟进维护高价值客户，驱动转介绍与复购业务增长。
AI 模型评估师（远程）
Outlier.ai
2024年10月 – 2025年6月
  • 设计并迭代大语言模型评估框架，从指令遵循、事实准确性、逻辑一致性与安全性等维度进行系统性质量评估，构建可复用评估方法论。
  • 对模型失效模式进行归因分析，输出结构化反馈支持模型迭代、数据清洗与 Prompt 优化。
  • 建立对抗性提示测试、边缘场景识别与评分量表设计的标准化文档体系，为评估团队提供可复用工作框架。
核心技能
AI / LLM
Prompt 工程、few-shot 学习、Workers AI、LLM 评估、Anthropic Claude API
云计算基础设施
Cloudflare Workers / KV / D1 / Pages，GCP，AWS，Linux (Ubuntu/Debian)，systemd
网络与安全
Shadowsocks，VLESS/XTLS-Reality，VMess，3X-UI，iptables，ufw，fail2ban，BBR，Let's Encrypt
编程语言
JavaScript (Node.js)，Python，R，SQL，Bash，SAS
数据分析
Power BI，Tableau，Streamlit，回归建模，CRM 分析
语言能力
英语（工作级流利）· 普通话（母语）