# Daniel Liu — hi-daniel.com

> Daniel Liu is a full-stack B2B marketer in San Francisco: growth,
> lifecycle, marketing operations and AI go-to-market, plus the engineering to
> build the systems those programmes run on. He owned a cross-border payments
> line from $2M to $250M+, ran a team of seven on a $10M budget, and shipped
> nine AI systems solo. This site is the portfolio, and every claim on it links
> to the case study behind it.

## Start here

- [Daniel Liu](https://hi-daniel.com/): Full-stack B2B marketer: paid, lifecycle, CRM and the operations layer under them, plus nine AI systems built solo and live on public URLs.
- [About Daniel Liu](https://hi-daniel.com/about.html): Ten years of B2B and fintech growth: paid media, web, lifecycle, ABM and marketing ops, plus the AI automation Daniel Liu builds on top of them.
- [Daniel Liu Resume](https://hi-daniel.com/resume.html): Daniel Liu's resume in six versions by role: performance, growth, lifecycle, AI GTM, marketing ops, head of marketing.
- [Proof: this site is the work sample](https://hi-daniel.com/proof.html): Every number this portfolio claims about itself, with the run conditions behind it: PageSpeed, accessibility, crawl health, and the decision behind each one.
- [Contact Daniel Liu](https://hi-daniel.com/contact.html): Hiring for AI-native growth, AI GTM or marketing ops?

## AI systems, each with a live public URL

- [AdFlow: Meta Campaign Decisions With Evidence](https://hi-daniel.com/ai/adflow.html): A prototype that runs one Meta account through the Marketing API.
- [AdGuard Ad Policy Checker](https://hi-daniel.com/ai/adguard.html): AdGuard is an AI compliance shield for ad creative.
- [AdRadar Competitor Advertising Intelligence](https://hi-daniel.com/ai/adradar.html): AdRadar collects one competitor's ads from LinkedIn, Meta and Google, labels the LinkedIn set ad by ad, and reads their strategy off those labels.
- [Career Capybara: Unified Job-Search Workspace](https://hi-daniel.com/ai/career-capybara.html): Career Capybara turns a job description into a tracked application, a tailored resume and interview prep, keeping the whole search in one workspace.
- [CreativeOS Policy-Aware Ad Copy Engine](https://hi-daniel.com/ai/creativeos.html): CreativeOS learns your brand voice from your own site and writes channel-ready ad copy for Google, Meta and LinkedIn that clears policy before you export.
- [Hyper Creator: Short-Video Content Hub](https://hi-daniel.com/ai/creator.html): A content hub for TikTok and Douyin: pull a creator's playlist, transcribe every video, translate it, rewrite it into your own script, then voice it.
- [FinGuard Fintech Marketing Compliance Co-Pilot](https://hi-daniel.com/ai/finguard.html): FinGuard is a fintech marketing compliance co-pilot: paste a draft, pick the jurisdiction, and get a severity-ranked audit against the real rulebooks.
- [MarketingOps: Email-to-Master-Data Assistant](https://hi-daniel.com/ai/marketingops.html): MarketingOps turns forwarded report emails into clean master data.
- [TechSpy Competitor Tech-Stack Intelligence](https://hi-daniel.com/ai/techspy.html): TechSpy fingerprints any competitor URL: HTML and JS bundles, DNS and email posture, and the real tech stack behind the marketing site.
- [TikTok Miner: TikTok Content Intelligence](https://hi-daniel.com/ai/tiktok-miner.html): TikTok Miner ingests a creator's whole catalog with its play counts, transcribes the videos you pick and embeds them into a corpus you can query.

## Case studies

- [Acquisition Efficiency: CPL $300 → $160](https://hi-daniel.com/work/acquisition-efficiency.html): Rebuilding a paid engine until every dollar produced a CAC finance could sign: LinkedIn enterprise CPL $300+ to ~$160 and Google registration $90+ to ~$35.
- [Agent-Native Content Engines](https://hi-daniel.com/work/agent-native-content-engines.html): Two agent-native content engines at ThinkingAI: a social calendar and a newsletter, each carried from plan to a near-ready deliverable by multi-LLM pipelines.
- [Agent-Native Reporting: The Self-Writing Week](https://hi-daniel.com/work/agent-native-reporting.html): An agent-native operating model for status reporting: nine channels compile my week, my manager's agent consolidates the team's, and the two agents talk.
- [AI Video Pipeline: Brief to Finished Film](https://hi-daniel.com/work/ai-video-pipeline.html): A two-track AI video pipeline: Kimi K3 for concepts, Claude directing, Seedance for cinematic mood, and a Remotion renderer for pixel-accurate product UI.
- [Capstone: Reusable Multi-Agent Systems](https://hi-daniel.com/work/capstone-agent-systems.html): One agent-team skeleton reused across six marketing functions, governed by a single rubric and hard fact-check gates, with a meta-optimizer over the top.
- [CMS Rebuild: Funnel-Aware Sanity Platform](https://hi-daniel.com/work/cms-rebuild.html): Designing and hand-coding a funnel-aware Sanity CMS for ThinkingAI: nine content types, structural funnel encoding, self-owned lead capture and attribution.
- [Global Web & Tracking: Multi-Region Lead Flow](https://hi-daniel.com/work/global-web.html): Website, CMS, tracking and SEO run as one governed operation across regions, so every market ships fast and every conversion path stays countable.
- [Marketing Ops: 200+ Workflows in Production](https://hi-daniel.com/work/marketing-ops.html): I rebuilt a manual GTM back office into 200+ automated workflows on one shared object model, one cleaned dataset behind 8 reports, lead handoff at 1-2 minutes.
- [Reports & Insights Portal: Self-Hosted BI](https://hi-daniel.com/work/reports-insights-portal.html): A self-hosted, auto-updating marketing BI portal: five data sources pulled through their own APIs, role-modular reports, refreshed daily by GitHub Actions.
- [Revenue Engine: $2M → $76M → $250M+](https://hi-daniel.com/work/revenue-engine.html): The revenue engine behind a new B2B payments line: instrumented acquisition, a lifecycle engine and the automation under both, from first campaign to scale.
- [AI Sales Intelligence: A Ranked, Reachable Book](https://hi-daniel.com/work/sales-intelligence.html): A pipeline that took 5,407 Ai4 companies to 944 ranked targets, built from cheap-LLM labeling, Apollo enrichment, and a report sales can work.
- [Web-Coded US Site: Shipping Pages Solo](https://hi-daniel.com/work/web-coded-us-site.html): Taking over the ThinkingAI US marketing site and moving the build to a local, AI-assisted web-coding pipeline with a written SOP, across the v2.1.9 to v2.5.1 chain.

## How the work is framed by role

- [AI GTM: Nine Agent Systems Built Solo](https://hi-daniel.com/lenses/ai-gtm.html): Nine AI GTM systems shipped solo, each on a live public URL: competitor intelligence, ad compliance, content mining and reporting automation.
- [AI-Native Growth: One Revenue Line to $250M+](https://hi-daniel.com/lenses/growth-lead.html): I ran the demand side of a B2B cross-border payments line at PingPong, from $2M at handover to $76M in year one and past $250M+ in year two.
- [Marketing Ops & Systems: 200+ Workflows Live](https://hi-daniel.com/lenses/revops-gtm.html): The ops layer under a $250M+ revenue line: CRM architecture, routing, attribution and reporting wired so sales, marketing and finance read the same number.

## Skills, end to end

- [Email & Lifecycle: Signup to Expansion](https://hi-daniel.com/expertise/email-lifecycle.html): Lifecycle cadences keyed to account state, guiding users from signup to activation, conversion and retention on Mailchimp, HubSpot and Salesforce Flow.
- [Growth Marketing: System-Driven Growth](https://hi-daniel.com/expertise/growth-marketing.html): Full-funnel growth for B2B and fintech across North America, Europe and Southeast Asia: acquisition, activation, retention and revenue, all instrumented.
- [Marketing Ops & Analytics: CRM and Attribution](https://hi-daniel.com/expertise/marketing-ops.html): CRM architecture, automated routing, attribution and reporting for B2B and fintech GTM teams, built on Salesforce, HubSpot and 200+ live workflows.
- [Partner & Affiliate: Channels That Compound](https://hi-daniel.com/expertise/partner-affiliate.html): Affiliate and referral programs from B2C home improvement to B2B fintech: commission models, partner enablement, attribution and fraud control.
- [SEO / GEO & Content: Organic Engines](https://hi-daniel.com/expertise/seo-content.html): Technical SEO, information architecture and scored editorial workflow for multi-region B2B and fintech funnels, built to rank in Google and get cited by AI.
- [User Acquisition: Tuned to CAC Payback](https://hi-daniel.com/expertise/user-acquisition.html): Multi-channel paid acquisition across Google, Meta, LinkedIn and programmatic, measured on CAC payback and lead quality channel by channel.

## Lifecycle principles

- [Activation: Triggered by Behavior](https://hi-daniel.com/principles/activation.html): Replacing a 30-day calendar drip with five account-state triggers, so new users activate on what they did rather than on how long they have waited.
- [Expansion: Cross-sell on Usage Signals](https://hi-daniel.com/principles/expansion.html): Building expansion campaigns on product-usage signals and lifecycle segmentation, so the next product is offered when account behavior says it is time.
- [Onboarding: Getting First Steps Done](https://hi-daniel.com/principles/onboarding.html): Onboarding is finished when the account is routed, verified and funded.
- [Retention: Respond Before the Customer Does](https://hi-daniel.com/principles/retention.html): A verification stalls, a document goes nowhere, usage quietly drops.

## How the AI work is architected

- [AI Systems & Agent Architecture](https://hi-daniel.com/ai.html): AI pays off when someone turns what a company knows into context agents can run on.
- [MCP Server](https://hi-daniel.com/mcp.html): A Model Context Protocol server over this portfolio.

## Writing

- [Marketing Ops, Lifecycle & AI GTM Notes](https://hi-daniel.com/blog.html): Playbooks, postmortems and working numbers from ten years of B2C affiliate and B2B fintech: marketing ops, lifecycle, paid acquisition, SEO and AI agent systems.

## Posts

- [Running website projects without due dates](https://hi-daniel.com/blog/a-website-roadmap-with-no-dates-on-the-cards.html): A sprint board for the people doing the work and a quarter map for everyone asking about it, both reading the same records so they cannot drift apart.
- [AI search optimization (GEO): what matters](https://hi-daniel.com/blog/ai-search-optimization-geo-what-actually-matters-now.html): How AI answer engines actually pick sources: unlinked mentions beat backlinks, extractable structure beats polished prose, and a robots.txt audit comes first.
- [Two listicles beat my whole site: GEO notes](https://hi-daniel.com/blog/ai-search-optimization-more-thought.html): Months of GEO tests on my own domain, read back slowly: what the log actually proved, what stayed unproven, and which claims I would still bet on.
- [Search Ads research as an agent pipeline](https://hi-daniel.com/blog/auditing-3156-competitor-paid-keywords.html): Search Ads competitor research rebuilt as an agent pipeline: DeepSeek cleans, Kimi K3 reasons, Claude sets strategy, the Google Ads API ships the campaign.
- [A referral-led growth system for B2B payments](https://hi-daniel.com/blog/building-a-referral-led-growth-system-for-a-new-b2b-payment-feature.html): Growing a B2B payments feature through existing customers: adoption first, then referrals into their partner networks, with email keyed to account state.
- [ABM by tech signals: 2% to 31% precision](https://hi-daniel.com/blog/building-an-abm-list-from-technology-signals.html): Building a North America ABM universe from technology signals instead of industry codes, and the version-by-version precision numbers behind each filter.
- [A referral lifecycle that keeps trust intact](https://hi-daniel.com/blog/designing-a-referral-driven-lifecycle-without-breaking-trust.html): Suppliers met the product through a payment.
- [Email Marketing: Timing Beats Copy](https://hi-daniel.com/blog/email-marketing-notes-from-the-field.html): What moves lifecycle numbers in B2B and fintech: timing, triggers and clean user state, with copy a distant fourth.
- [A Two-Month Tracking Bug With a One-Hour Fix](https://hi-daniel.com/blog/fixing-a-cross-domain-tracking-problem-that-shouldnt-have-taken-two-months.html): Paid attribution was dead for two months and every UTM evaporated at the domain crossing.
- [Launching a B2B fintech affiliate program](https://hi-daniel.com/blog/from-b2c-affiliate-experience-to-launching-a-b2b-fintech-affiliate-program.html): What years of B2C affiliate buying taught me, and the harder lessons B2B fintech added: partner vetting, payout structure, attribution and fraud control.
- [Landing page testing without overthinking it](https://hi-daniel.com/blog/how-i-approach-landing-page-testing-without-overthinking-it.html): Most landing pages break the promise the ad made.
- [Automating a messy monthly report with an LLM](https://hi-daniel.com/blog/how-i-automated-a-messy-monthly-report-using-llm.html): A raw HQ export, a Zapier cleaning flow and a custom GPT: turning a two-hour monthly close into five minutes, with PII stripped before any model sees a row.
- [Marketing ops that replaced chaos with a queue](https://hi-daniel.com/blog/how-i-built-a-marketing-operations-processes-that-reduced-chaos-protected-focus-and-made-scaling-possible.html): The team executed well and still drowned, because every request arrived as an interruption.
- [An AI lead scoring assistant for B2B fintech](https://hi-daniel.com/blog/how-i-built-an-ai-lead-scoring-assistant-that-actually-works-in-b2b-fintech.html): An interpretive lead qualification system with no engineers and no black box, wired through the existing stack, scoring one question at a time.
- [Email intake rebuilt: one form, one board](https://hi-daniel.com/blog/how-i-turned-email-chaos-into-a-calm-scalable-marketing-operation.html): Separating one-off email requests from living cadences, enforcing one intake door, and building the launch sequence that made sends predictable.
- [ChatGPT for SEO: intent, metadata, audits](https://hi-daniel.com/blog/how-i-use-chatgpt-to-optimize-content-for-real-seo-gains.html): ChatGPT researches and edits; I write the page.
- [Two AI flows for a one-person support desk](https://hi-daniel.com/blog/how-i-used-simple-automation-ai-to-help-our-support-team.html): One rep, one inbox, four departments' mail.
- [Fintech keywords: clicks, compliance, intent](https://hi-daniel.com/blog/keyword-research-for-fintech-company.html): Positions held, impressions held, clicks fell.
- [Onboarding email rebuild: KYC time 140h to 48h](https://hi-daniel.com/blog/rebuilding-our-onboarding-email-system.html): One shared MailChimp account, three regions, no owner.
- [Managing trade shows from one page per show](https://hi-daniel.com/blog/running-a-trade-show-programme-from-one-page.html): Every show as one addressable workspace with the same five sections, plus a reusable layer that makes the next show cheaper than the last.
- [Running notes on GEO and AI search](https://hi-daniel.com/blog/running-notes-on-geo-ai-search-and-what-seems-to-actually-matter.html): Verifying an hour of GEO advice one claim at a time against my own site: make the change, watch the log, record what the engines actually did.
- [Running social solo: a $20 test loop](https://hi-daniel.com/blog/running-social-campaign-with-limited-resources.html): Running social as a system when you cannot outspend anyone: one message library, reusable templates, a weekly rhythm, and a cheap creative test loop.
- [Scaling a product launch on record design](https://hi-daniel.com/blog/scaling-a-new-product-launch-with-marketing-ops-infrastructure.html): The biggest risk in a referral-driven launch is data.
- [Slicing a whitepaper into a lead engine](https://hi-daniel.com/blog/slicing-a-14-chapter-whitepaper-into-a-lead-engine.html): The usual move is one form in front of one PDF.
- [The trade-show follow-up system: badges to CRM](https://hi-daniel.com/blog/the-trade-show-follow-up-system-badges-to-crm.html): Most trade-show leads die in a pocket.
- [Data + AI to write articles that rank](https://hi-daniel.com/blog/use-data-ai-to-help-my-content-team-write-articles-that-rank.html): A scored content pipeline for a team writing in a second language: competitor scans, an AI outline, and a measurable score before a word gets drafted.

## Querying this site directly

There is a Model Context Protocol server over the whole portfolio at
`https://hi-daniel.com/api/mcp` (Streamable HTTP, JSON-RPC 2.0, no auth). It
exposes `search_portfolio`, `get_page`, `list_pages` and `get_profile`, so an
agent can answer questions about this work without crawling. Documentation:
https://hi-daniel.com/mcp.html

## Notes for indexing

- Every page is server-rendered static HTML. No content requires JavaScript.
- Canonical host is `hi-daniel.com`; `www` 301s to it.
- Sitemap: https://hi-daniel.com/sitemap.xml
- GPTBot, ClaudeBot, PerplexityBot and Google-Extended are explicitly allowed
  in robots.txt.
- Content may be quoted and cited with attribution to Daniel Liu and a link to
  the source page. Figures on this site are from real programmes; please do not
  restate them without the link.