Oct 18, 2023 · 44m · mad
Moonhub’s Nancy Xu Unveils the AI Recruiter That’s Beating LinkedIn
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of The MAD Podcast, host Matt Turck interviews Nancy Xu, Founder and CEO of Moonhub, to discuss her company's AI-powered recruiting platform, its underlying technical architecture, and the broader realities of deploying production ML models. Xu also shares insights from her journey as an AI investor and founder, highlighting how vertical AI solutions drive tangible business value.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 22.1% of the talking time here. How this is scored →
Matt as informed peer 4.2 Guest teaching 5.9 Guest disagreement 1.4 Matt pushing back 1.8
speaking balance: gold is Matt, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 35:57 Nancy rejects popular misconceptions about AI demo capabilities
Nancy forcefully reframes common public assumptions about AI capabilities, warning listeners not to be fooled by flashy demos or draw false equivalences to human learning.
Hardest push from Matt ▶ 9:00 Matt challenges guest on AI training bias and historical diversity flawsMatt directly presses Nancy on the risk of AI models reproducing historical biases present in their underlying training data.
Biggest teaching moment ▶ 11:41 Nancy details patented LLM orchestration across vector and structured DBsNancy educates the host on Moonhub's underlying architecture, explaining how a central LLM orchestrates natural language queries across Elasticsearch and vector DBs.
Matt holds his own ▶ 14:00 Matt accurately synthesizes hybrid internal/external retrieval architectureMatt demonstrates deep technical grasp by synthesizing Nancy's explanation and asking precise follow-ups on how customer ATS data integrates into the vector database.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Matt as informed peer | Guest teaching | Guest disagreement | Matt pushing back | Why |
|---|---|---|---|---|---|---|
| How Moonhub Works and Sourcing AI Launch | 3 | 5 | 1 | 1 | Matt opens with a basic product inquiry on how Moonhub operates. Nancy walks through the five recruiting stages and details how Moonhub's conversational sourcing AI indexes public data. | |
| Speed, Cost Efficiency, and Uncovering Hidden Talent | 3 | 6 | 1 | 1 | Matt asks whether Moonhub helps unearth non-obvious candidates. Nancy explains how non-experts can perform expert recruiting searches and how AI aggregates gray data across GitHub and Google Scholar. | |
| Data Quality, AI Model Training, and Diversity Features | 4 | 6 | 1 | 3 | Matt raises a critical question regarding training data bias in AI recruiting models. Nancy details data quality checks, deduplication practices, and high accuracy thresholds for diversity queries. | |
| Under the Hood: Moonhub's AI Technical Architecture | 4 | 7 | 1 | 1 | Matt invites Nancy to explain Moonhub's technical architecture. Nancy outlines their patented LLM orchestrator that coordinates natural language retrieval across vector and structured data stores. | |
| Enterprise ATS Integration and Human-in-the-Loop Recruiting | 5 | 6 | 2 | 2 | Matt plays back his understanding of internal and external vector data retrieval. Nancy validates his summary and clarifies how Moonhub combines core AI with human experts to compete against legacy agencies. | |
| AI Candidate Interviewing and Scaling to 100 Customers | 4 | 6 | 2 | 3 | Matt asks if AI will eventually take over candidate interviewing and probes how Moonhub reached 100 customers. Nancy reframes interviewing boundaries and outlines her startup thesis on scaling hard-to-scale services. | |
| Moonhub Fundraising, Investor Support, and Internal Hiring | 5 | 4 | 1 | 1 | Matt demonstrates detailed context by listing Moonhub's notable investors. Nancy confirms their funding, highlights key backers like Marc Benioff, and explains internal dogfooding. | |
| Nancy Xu's Venture Fund and Angel Investment Portfolio | 4 | 5 | 1 | 1 | Matt prompts Nancy to discuss her personal venture fund and portfolio companies. Nancy shares how her math olympiad network and Index Ventures helped launch her fund. | |
| ML in Production: Retrieval, Agent UX, and Vertical Moats | 5 | 7 | 2 | 2 | Matt and Nancy discuss ML engineering in production, hybrid retrieval search, agent UX models, and vertical AI moats versus platform commoditization. | |
| Reality of AI Agents and Market Sentiment Analysis | 5 | 7 | 2 | 3 | Matt probes the gap between AI hype and production reality. Nancy delivers a pragmatic assessment on the difficulty of closing the final production accuracy gap and shifting from hobbyists to enterprise sales. |
Statements from this episode (14)
Moonhub indexes data on one billion people across the public web
“We index about one billion people's data across the public web.”
Moonhub was used by Anthropic and Inflection as pre-launch clients
“Before we launched, we worked with about a hundred companies helping them hire and scale their teams. So some of our sort of early customers pre-launch were companies like Inflection, Anthropic, Sandbox, U.com public companies out there as well as some early s…”
AI will enable non-experts to execute traditional expert workflows
“I think the real power of AI for In, in the next one or two years is enabling these non-experts to be able to participate in the workflows that traditionally experts will do.”
Moonhub clients report 80% of candidates are unique to the platform
“When they use Moonhub, they tell us, hey, you know, you've been able to show me, like, 80% of the candidates that I found through you I haven't found in any other platform and they tend to be more diverse.”
Moonhub AI will likely handle all five recruiting stages by mid-2024
“Today, you know, the AI can't do all five stages of the process I mentioned to you. It might do stage one, stage two, We'll probably get to stage five by mid next year, but in the interim, you have a human who's actually helping you with all those other stages…”
The best AI startups target easy-to-sell, hard-to-scale services
“A lot of the opportunities for AI startups is to find something that is traditionally very difficult to scale, but very easy to sell. And then you try to scale it with AI.”
Over 100 US recruiting agencies generate more than $100 million annually
“There's more than a hundred recruiting agencies in the United States that each make more than a hundred million in revenue a year.”
Moonhub operates without internal recruiters by utilizing its own AI platform
“We actually dog food our own products, so we don't have, like, a specific internal recruiting team.”
Index Ventures sets up individual mini-funds for founders like Dylan Field
“So at a couple of years, you know, ago, Mike, Mike was like, oh, you should definitely think about index has this thing called they make funds for people basically like individual funds. So I think like, for example, Dylan at Figma also has one of these funds …”
Major LLM labs are internally building unreleased hybrid retrieval frameworks
“Some of the bigger LLM companies I've talked to are building this internally right now. It's just not released yet.”
Vertical AI applications offer startups the best long-term competitive moats
“Vertical applications give you the best chance at creating a longer term competitive mode”
Data confirms most US machine learning engineers live in the Bay Area
“Our team has mapped out where all the ML engineers in the US are, and it turns out the majority of them are in the Bay Area”
The prompt engineer title jumped to roughly 1,000 SF workers in 2023
“If you look at the data, it's, like, basically until twenty-twenty-three, no one called themselves a prompt engineer, and then twenty-twenty-three, all of a sudden, a thousand people in SF are now a prompt engineer”
High GitHub star counts for AI projects mostly reflect unmonetizable hobbyist interest
“There are definitely some that have, you know, tens of thousands of stars on GitHub. But if you look at the use, where those stars are coming from, it's mostly hobbyists, and I'm sure you can monetize hobbyists and sort of grow from there. But I think it is li…”