 | Kelvin Lau (Luu)Systems & Revenue. Waterloo CS. |
Bridging software architecture and enterprise revenue. Waterloo CS graduate with enterprise engineering roots (Amex, SMART Technologies) and commercial GTM execution (Aftersell). Building autonomous systems and scaling technical pipeline.
Bio
I graduated with a Bachelor of Computer Science (Business Option) from the University of Waterloo, backed by enterprise software engineering co-ops at American Express and SMART Technologies.
I then moved into commercial execution, from transactional e-commerce into high-ticket, consultative deals focused on ROI, cash flow, and capital allocation. Along the way I scaled a marketing agency for DTC brands, learned B2B distribution for wholesale clients, and built automated GTM infrastructure.
Today that crossover is the product: I write code to build revenue systems a traditional seller can't ship, and I close deals with commercial judgment a traditional engineer rarely develops. Whether that looks like a Mid-Market SE/AE seat or a 1099 engagement helping Series A-D SaaS teams, the same stack applies: AI automation, modern workflows, and pipeline that converts.
Case Studies
| 01 | Autonomous B2B Lead Generation & Waterfall Enrichment Pipeline The Enterprise GTM & Waterfall Enrichment Engine. Built for a Red Light Therapy system client in Boston, MA (in partnership with a managing partner and ex-owner of Suncapsule). How it works Ingestion: Processes raw Outscraper gym data for targeted U.S. states. Filtering & Tiering: Scripts built via Cursor segment accounts and score independent operators into fit tiers based on website indicators. Waterfall Enrichment: Passes qualified data to Clay and Instantly to automate cold outreach at scale. CursorOutscraperClayInstantlyPythonSerper [View Architecture & Code on Github] |
| 02 | Autonomous Multi-Region GTM Engine & Conversion Optimization An enterprise-grade outbound system built on n8n, Apify, and Airtable to automate lead discovery, fit-scoring, and multi-persona outreach across global markets. The Objective & Problem Solving distribution friction for enterprise sales operations. Standard outreach pipelines suffer from low conversion velocity, high token costs from unnecessary LLM generation, and application fatigue. Architectural Evolution (V1 vs. V2) The V1 Bottleneck: Built dynamic LLM resume-tailoring per job posting. Live testing across ~340 sends revealed a ROI bottleneck: heavy Google AI token burn, with minimal lift in enterprise conversion rates versus static master resumes. The V2 Optimization: Demoted the LLM from content generation to high-precision classification (Job Fit Scoring). Automated routing of deterministic Master Resumes based on target persona (SE vs. AE). Human-in-the-Loop Orchestration: Generated automated, dynamic LinkedIn outreach scripts for Hiring Managers and Peer ICs directly in the database, reserving manual high-touch execution strictly for Tier 1 targets. Architectural Exploration (V3 Vision vs. Pragmatic Scope) The V3 Exploration: Explored expanding the pipeline into a fully autonomous ATS auto-apply agent using Playwright/Browserless to parse Ashby, Lever, and Greenhouse forms and execute dynamic LLM screening answers. The Pragmatic Tradeoff (Why V3 Was Scoped to Human-in-the-Loop): During technical validation, automated session-handling revealed a critical security risk:
Decision: Intentionally bound automation to Data Ingestion, Persona Deduction, and Message Staging. Preserved a 10-second manual execution step for form submission and outreach, delivering 90% of full-automation velocity with 100% account security and zero system downtime. Deterministic Execution: Runs daily at 8:00 AM EST across APAC (HK/SGP), Canada, and USA markets. Resilience & Infrastructure: Self-hosted on PikaPods with batch timer throttling, rate-limit handling, and error-handling fallbacks. n8nApifyGoogle AI Studio (Gemini)AirtablePikaPodsInstantlyProspeoTelegramGoogle DrivePDFShift [View n8n Workflow JSON & Architecture on GitHub] |
| 03 | AI-Agent Pipeline & Revenue Recovery Engine The Autonomous Pipeline Generator · In Progress. Built for Series A-D SaaS and capital equipment manufacturers that leak pipeline daily: inbound signals go cold, SDRs burn hours on manual research, and high-intent accounts sit unattended. Target outcomes: recover unworked opportunities before competitors touch them, and cut manual SDR overhead by automating signal → research → personalized outreach. How it works Signal Intake: RSS / funding read-pulls surface net-new buying events (raises, hiring spikes, product launches) as soon as they hit the wire. Bottleneck Agents: AI agents score accounts, infer operational bottlenecks, and rank who is most likely to buy, so reps only touch Tier-1 recovery work. Revenue Recovery Push: Hyper-personalized outreach sequences land in Clay/n8n with low-overhead 1099/contract execution attached, capturing pipeline that would otherwise expire unworked. n8nAI AgentsRSS FeedsClay |