01
Talent Lead Exercise · Confidential

Runlayer's
Talent
Exercise

Prepared by Priscilla Philavong  ·  May 2026
Please enter your first and last name.
Press Enter ↵ to continue
Non-Disclosure Agreement

NON-DISCLOSURE AND CONFIDENTIALITY AGREEMENT

Unilateral · Candidate Materials
Effective Date: May 26, 2026
Disclosing Party: Priscilla Philavong, New York, NY
Receiving Party: [ You ], authorized representative of Runlayer, Inc., 31 Bond St, Floor 5, New York, NY 10012

The Receiving Party wishes to review confidential candidate materials prepared by Priscilla in connection with the Talent Lead interview process at Runlayer, Inc. (the "Purpose"). In consideration of this disclosure, the Receiving Party agrees as follows:

1 · Confidential Information

All materials, strategies, frameworks, candidate assessments, employer brand frameworks, AI workflow descriptions, sourcing methodologies, and any work product contained in or derived from this Talent Lead Exercise, whether disclosed in written, electronic, or verbal form, constitute Confidential Information.

Excluded: information the Receiving Party lawfully possessed beforehand; information publicly available through no fault of the Receiving Party; disclosures required by law.

2 · Limited Use

The Receiving Party shall use Confidential Information solely for evaluating Priscilla's candidacy and for no other purpose. The Receiving Party shall not reproduce, distribute, publish, or disclose any portion thereof to any third party without Priscilla's prior express written consent. This obligation remains in full force indefinitely.

3 · Ownership

Nothing herein conveys any right, title, interest, or license in the Confidential Information. All strategies, frameworks, methodologies, and intellectual property remain the sole and exclusive property of Priscilla Philavong.

4 · Return of Materials

Upon conclusion of the interview process or upon Priscilla's written request, the Receiving Party shall promptly destroy or return all materials containing Confidential Information.

5 · Remedies

The Receiving Party acknowledges that a breach may cause irreparable harm for which monetary damages are insufficient. Priscilla is entitled to seek injunctive relief and any other available remedies without proving actual damages or posting bond.

6 · Electronic Signature

The Receiving Party's electronic signature shall be deemed valid and legally binding pursuant to the E-SIGN Act and UETA. This Agreement is governed by the laws of the State of New York.

Electronic Signature *
Sign here

Exercise Responses

Enter the access code to view the exercise responses.

Incorrect access code. Please try again.
🔒 Confidential · Acknowledged by ·
RUNLAYER · Talent Lead Exercise · Confidential

Priscilla Philavong

priscillaphilavong@gmail.com  ·  +1 917.344.9379  ·  May 2026
Section A
Talent Taste — Product Engineer

1. Three Candidate Profiles

Junior 1–2 yrs  ·  Mid 2–4 yrs  ·  Senior 4+ yrs in production AI/LLM/MCP environments

Senior · 4+ yrs production AI/LLM/MCP
Eric Anderson — AI Product Engineering Lead, Superblocks (3.25 yrs) → AI Product Engineer, Endgame · Brooklyn, NY

Why this profile excites me

Superblocks' product governs apps centrally with integrations, authentication, permissions, and audit logs in a single pane of glass. Eric led the AI product engineering that built it. Near-identical architecture to Runlayer. Not adjacent to the problem — has been solving it for 3+ years.

Yellow / red flags

  • Started at Endgame 7 months ago — understand why before reaching out
  • Moderate LinkedIn interest signal — approach matters

Confirm early

  • How much of the auth and permissions layer did you own end-to-end vs. contribute to?
  • What did customer interaction look like — direct discovery or specs handed to you?
  • Why Endgame, and what's pulling you to consider something new?

Sourcing — small pool, network mapping required

Zapier MCP alumni · Anthropic solutions · OpenAI enterprise tooling · LangChain founding engineers · NYC: Oasis Security (sleeper pick — secures API keys/OAuth tokens, near-identical problem space), Apiiro, Tines, Vanta

Mid-level · 2–4 years
Daniel Kimmelmann — Solutions Lead, Postman Flows (3.7 yrs) · NYC · Open to work

Why this profile excites me

Built Postman Flows — their AI agent + API workflow product — from pre-beta to GA. Primary SME for all escalated enterprise customer issues. Built ~100 runnable examples, delivered live conference demos, and used Flows internally to automate IT, HR, Recruiting, Marketing, and Sales. The Runlayer PE JD asks for someone who writes production code AND works directly with customers. Daniel has been doing exactly that for 3.7 years.

Yellow / red flags

  • Title is Solutions Lead, not Software Engineer — coding depth needs validation
  • Difference between configuring workflows and writing production Python/TypeScript is real

Confirm early

  • Walk me through something you built in Flows that ended up in production. How much was you writing code vs. configuring the product?
  • What's the hardest API integration you've debugged and what broke?
Junior / Stretch · 1–2 years
Melissa Allan — Applied AI Engineer, Workbench AI (9 mos) · NYC · Open to work

Why this profile excites me

Stack: LangGraph, multi-agent orchestration, Claude, FastAPI, Pydantic, Pinecone, RAG. Not using AI tools — building agentic systems from scratch with typed contracts and explicit error handling. Prior VMware PM background gives product instincts most engineers at this level lack.

Note on JD tension

The JD asks for 4+ years. Deliberate stretch bet — align with Tal before opening this pipeline. Direct LangGraph + Claude experience closes a meaningful portion of the gap that years alone can't.

Yellow / red flags

  • Workbench AI is a self-directed project — no enterprise customer experience
  • Role requires working directly with Gusto, Instacart, Ramp

Confirm early

  • Has anyone paid for or operationally depended on what you built?
  • What's the hardest failure you debugged in a multi-agent pipeline?
  • How do you think about building systems an enterprise security team would trust?

Candidate Archetypes — 3 Sourcing Buckets

1. Production AI/integration engineers from developer tools companies — Zapier, Retool, Stripe, Cloudflare, Twilio, Postman, Superblocks. Customer-facing, enterprise API depth, translate pain into product. Maps directly to JD's Salesforce/ServiceNow/Jira/Slack callouts.

2. AI solutions engineers from LLM companies — OpenAI solutions, Anthropic tooling, Langchain, Modal, Cursor (cursor.com, the AI code editor). MCP ecosystem fluency natively. Hardest bucket to find, highest signal when you do.

3. NYC-based security and compliance engineersOasis Security (sleeper pick — secures API keys/OAuth tokens, near-identical to what Runlayer governs for MCP), Vanta, Drata, Tines, Apiiro, Adaptive Security, Stytch, Dashlane, DataRobot. Enterprise security buyers, integration-heavy, NYC culture compatible.

Broad

Zapier · Retool · Stripe · Cloudflare · Twilio · Modal · Langchain · Anthropic · OpenAI · Cursor (cursor.com) · Replicate · Vercel · Datadog · Wiz · Snyk · DataRobot · Antimetal · Postman · Superblocks · EliseAI

NYC-Based

Oasis Security · Vanta · Drata · Tines · Apiiro · Adaptive Security · Stytch · Hebbia · Hex · Assembled · Sublime Security · DataDome · Actively AI · DataRobot · Butter · Harbor IT


2. Startups with High Technical Talent Density

Cursor (cursor.com) · Early stage

~40 people, hundreds of millions in ARR. Ex-OpenAI team, full ownership, fast shipping. Engineers use and understand MCP clients daily — Cursor is one of 300+ clients Runlayer supports.

Linear · Early/growth

Small team, extreme craft standards. Every engineer ships end-to-end without a PM buffer. Product quality directly reflects team quality.

Wiz · Growth stage

$0 to $100M ARR in 18 months. Unit 8200 alumni — engineers who've operated in high-stakes security environments. Directly relevant talent pool for Runlayer.

Databricks · Growth stage

Strong ML and platform engineering culture. Existing network at Runlayer through GTM leadership — fast path to warm introductions.

Key ingredients at Series A: Technical founders personally in the first 20 hires. Narrow sourcing — 15 right people, not 500 blasts. Technical employer brand in open source and conference talks before the careers page. A fast, respectful process that signals company quality. Equity that rewards early risk.
Section B
Employer Brand — 3 Core Narratives + EVP
1 · "We helped build the protocol. Now we're securing it."

Uniquely Runlayer's — no competitor can claim it. The founding team built the first MCP server at Zapier. David Soria Parra (MCP creator at Anthropic) is on the cap table. Travis McPeak (Head of Security at Cursor) is an advisor. Every employer brand touchpoint reinforces this provenance.

Activation

Founding team publishes first-person pieces about what they saw at Zapier — the blind spots, the attack vectors. Engineers want to work with people who've been inside the problem.

2 · "Runners SHIP — full ownership, no bureaucracy, real stakes."

The people identity (Runners), values (Say it, Ship it, Scale it), and operating principle of intellectual humility point to the same place: we hire exceptional people and get out of their way. 9 products shipped in 8 months. What you build runs at Gusto, Instacart, and Ramp.

Activation

Engineering blog posts from Runners about what they shipped, how fast, and what they learned. The brand is what Runners say about their work — not what Runlayer says about itself.

3 · "The infrastructure moment."

MCP is the connective tissue of the agentic web. Runlayer is the security layer for it. For engineers who think in decades: the infrastructure you build here will be foundational to how AI operates in enterprise for the next 10 years.

Activation

Conference presence. The founding team already speaks internationally on MCP security. Get them on stage at AI infrastructure and security conferences — the engineers in those rooms are exactly who Runlayer needs.


The EVP Framework — Runners SHIP

S
Say it
Direct, not diplomatic. Build trust through honesty.
H
Humility
Know what you know. Name what you don't. Close the gap fast.
I
Impact
Shipping is the eval loop. Your work runs at Gusto and Ramp.
P
Protocol
We are the trust layer — for the product and for each other.

Give

Full ownership from problem to delivery. No PM buffer, no committee, no permission needed. You work inside the stack you're securing.

Get

Direct access to the people who built MCP. 5M+ secured calls. Real customers who depend on what you shipped. AI fluency as the default.

"We hire exceptional people and then get out of their way. Runners own their domain from problem to delivery — no handoffs, no bureaucracy, real stakes. If you want full ownership of hard problems at a company defining a category in real time, this is where Runners belong."
Section C
AI in My Workflow

Top 3 ways I currently use AI

1 · RPS Summarizer (ChatGPT + Brighthire) — used daily

Takes Brighthire transcripts from recruiter phone screens and converts them into structured hiring manager review docs automatically: takeaway, summary, strengths/weaknesses, and role alignment cross-referenced against the career ladder, JD, salary budget, and market benchmark comp by geo, size, and industry. Auto-submitted in Ashby, triggers HM notification immediately. Zero admin, zero delay between screen and HM review.

2 · Outbound Sourcer Agent (Claude)

Takes a candidate profile — GitHub, conference talks, open-source contributions, LinkedIn — and generates personalized outreach referencing their specific work. Adapts by persona and seniority.
Result: 46.6% InMail response rate over 6 months — nearly 2x the LinkedIn benchmark of ~25%.

3 · TA Calibration Agent (ChatGPT)

Synthesizes all panel feedback, flags scoring inconsistencies, surfaces unsupported opinions, produces a hire/no-hire recommendation. Separates "signal" from "confidence."
Result: Debrief conversations are faster and more honest — panelists arrive having seen where they diverge, not ratifying a consensus formed before the meeting.


Full AI recruiting stack — 5 agents in production

RPS Summarizer

Brighthire → structured HM review → auto-submitted in Ashby → HM notification. Full loop, zero admin. Includes comp alignment against market benchmarks.

Outbound Sourcer Agent

Personalized outreach by persona and seniority. 46.6% InMail response — 2x LinkedIn benchmark. Built on Claude.

TA Calibration Agent

Synthesizes panel feedback, flags inconsistencies, surfaces unsupported opinions, produces hire/no-hire rec.

Talent Brand GPT

EVP and recruiting messaging copilot. Persona modes, objection handling, voice consistency.

Offer Builder

Generates offer packets, comp framing, level alignment summaries, and closing narratives. Reduces offer-stage admin, directly improves close rates.


3 automations I'd build for Runlayer

1 · Pipeline Health Agent

Weekly digest: candidates gone cold, at risk of losing to a competitor, ready for an offer push. CRM intelligence applied to the candidate pipeline.

2 · Competitive Offer Intelligence Agent

When a candidate mentions a competing offer: surfaces public comp data, candidate motivations, and draft talking points for the close. Especially important at seed stage competing on mission and equity over cash.

3 · RPS Summarizer rebuilt for Runlayer's stack

The same Brighthire → Ashby loop built at Abnormal, adapted to Runlayer's interview process. Auto-submits structured recruiter screen summaries with role alignment against Runlayer's comp benchmarks and career ladder.

On using Runlayer's product in my recruiting workflow: I use MCP-connected tools daily — Claude Code, Glean, ATS integrations. I'd use Runlayer to set fine-grained permissions so AI recruiting agents can access candidate data but not comp data; audit every AI touchpoint with sensitive PII (GDPR/CCPA); and manage which MCP servers my recruiting stack is authorized to connect to. I can walk through this live using my actual stack.