AI Product Manager · The Pipeline Group

I turn the thing everyone works around into the thing everyone uses.

I'm Yonatan Nudman. I originated, prototyped, and shipped the three internal systems that The Pipeline Group's 200-person sales organization now runs on: ProspectLib, the company-wide outbound platform; Recruiting Funnel, the applicant tracking system; and Task Hub, which replaced roughly 30 Asana enterprise licenses.

Neuroscience→SDR→built the tool→the company built the role around it
Read the case studies → Download résumé ↓
~400%
lift in company-wide email volume after the platform I originated rolled out
200
SDRs on a platform that started as my single-rep prototype
~30
Asana enterprise licenses eliminated, thousands of dollars a month
3
internal products in production, scoped and prototyped solo
02 The arc

How a 500-call-a-day SDR ended up owning the roadmap.

Nobody handed me a product role. I built something the floor needed, it worked, and the job got written around it.

Oct 2024 — Mar 2025

Yellowbrick

Neuroscience Associate running neuromodulatory protocol research. Learned to design an experiment and read a result honestly.

Mar 2025 — Mar 2026

SDR, cybersecurity

Menlo Security, Cobalt.io, Island.io. A 500-call-per-day floor against a quota of 9 meetings and 5 qualified opportunities a month. Hit it consistently.

The inflection

I automated my own job

Built the first email-send automation for myself, then provisioned it to 30 reps. Team email volume jumped ~400%. That number made the case for a dedicated AI builder.

Mar 2026 — now

AI Product Manager

Promoted into a role that did not exist before. I scope, prototype, and drive to production the internal tools the whole sales org depends on.

03 The work

Three products, prototyped solo, productized with engineering.

Each one started the same way: I found a workflow everybody was quietly routing around, built the smallest thing that fixed it, and got the room to agree before a single engineer opened a ticket.

CASE 01 Live · 200 SDRs PRs #1916–#1918

ProspectLib — the outbound platform

The problem
Every SDR was composing outbound one message at a time. The ceiling on pipeline was not talent or targeting, it was keystrokes per rep per day. Nobody had framed it that way, so nobody had fixed it.
What I decided
Prove it at n=1 before asking for a single engineering hour. I built the automation for my own desk, measured the lift, then provisioned it to 30 reps to show it held at scale. Only once the ~400% number was real did I make the case to productize.
How it shipped
Partnered with LaunchQ engineering to rebuild the prototype as a native platform for all 200 SDRs. I seeded 42 Mirage Security templates so the rollout landed with real content instead of an empty shell, ran UAT, and merged PRs #1916 through #1918.
CASE 02 Shipped 9-stage pipeline

Recruiting Funnel — an ATS to replace the spreadsheet

The problem
Recruiting was running the entire hiring pipeline out of manual Google Sheets. Candidate status lived in whoever's head last touched the row, follow-up emails were hand-written, and there was no way to answer "where are we actually losing people?"
What I decided
Do the requirements work with the person who owns the pain, not around her. I drove definition directly with TPG's Head of Recruiting, Janice Hosea, which is why the stage model is configurable rather than my guess at what hiring looks like.
How it shipped
Scoped and prototyped it, then partnered with TPG engineering (Tiago and Seth) to build. Landed with 9 configurable pipeline stages, automated stage-transition emails so candidates stop falling through the cracks, and a 7-chart analytics dashboard that finally answers the drop-off question.
CASE 03 Shipped PR #1831

Task Hub — the product that paid for itself

The problem
Work was scattered across an expensive third-party tool and three unrelated habits: support requests, onboarding projects, and personal to-dos. We were paying ~30 Asana enterprise seats for something that only ever covered part of the job.
What I decided
Not "clone Asana." I defined the product around the three real workloads the org actually had — Support Tasks, Onboarding Projects, Personal Tasks — and built it inside LaunchQ, where the data already lived, so it inherited context instead of duplicating it.
How it shipped
Drove product definition with the manager stakeholders whose teams would live in it, led UAT and triaged real bugs from new feature requests, and shipped with engineering via PR #1831.
04 Also built

Things I shipped when nobody asked me to.

A PM who can build the prototype argues from a working thing instead of a slide. These are where that came from.

Commissions App
Google Apps Script web app that runs TPG's monthly xDR quota approval workflow in production. Finance signs off inside the tool.
TPG · in production
TPG-ASSISTANT
Natural-language Slack bot giving managers live Salesforce access in plain English. Meetings booked, rep leaderboards, quota pacing. Adopted by 20 managers.
TPG · 20 managers
Nomads AG stack
Freelance: a 3,065-line Discord-native leads bot on GoHighLevel webhooks with round-robin setter assignment and 5-checkpoint quality scoring, a Meta Ads analytics pipeline, and a self-healing monitor over 9 automations. Agency-reported: ~70% less manual CRM work, +25% lead-to-opportunity conversion, −18% CPL.
freelance · 24/7 prod
Wheelbot
Autonomous Python options bot running PMCC, Wheel, and VRP spread strategies. Alpaca API, SQLite in WAL mode, VIX-based position gating, daily-loss circuit breaker.
paper-trading live
Employee Lifecycle App
Next.js 16 prototype consolidating an 8-stage employee journey for TPG's People Pulse system.
employee-lifecycle.vercel.app ↗
05 How I work

The loop, in four moves.

Not a framework I read about. It is what actually happened three times in a row.

01Find the thing people route around

The best problems are already solved badly by hand. I go looking for the workaround, not the feature request, because the workaround tells you what the job actually is.

02Build the smallest working version

I prototype it myself before asking for engineering time. A working thing ends the debate that a spec would have started, and it de-risks the estimate.

03Get the room to agree first

I walk it past the CPO and the leads and answer every "what about" before handoff. Engineering should be building, not guessing at intent.

04Own it through UAT and after

I run the testing, sort real bugs from new feature requests, ship, tell people what changed and why, then watch how it gets used and cut what nobody touches.

06 Facts

Education

Bellevue University — Master of Business Administration
Expected August 2026 · GPA 4.0

Based

New York / Chicago · remote-first
Open to PM, APM, and AI PM roles

Product

Discovery · scoping · prototyping · stakeholder alignment · UAT ownership · roadmap planning · cross-functional partnership with engineering

AI & LLM

OpenRouterClaudeGeminiGrokprompt engineeringagent orchestrationAI reply classificationRAG

Builds with

TypeScriptJavaScriptPythonC#SQLNode.jsNext.jsReact.NET 8 BlazorApps ScriptExpressDiscord.js

GTM & infra

Salesforce (SOAP/REST/SOQL)HubSpotGoHighLevelZoomInfoVercelAzure SQLSupabaseDigitalOceanGitHub Actions

If something here sounds like a problem you have, let's talk.

Email
Nudman.Yonatan@gmail.com →
LinkedIn
/in/yonatan-nudman ↗
Résumé
One page, PDF ↓