Juan Pablo Pacheco

AI Solutions Engineer

I turn manual sales, customer and operations work into AI agents and workflows that run in production.

15+

years in enterprise sales, a quota in every role

1,000+

active users on 4 continents at JoGo.Team

80%+

of at-risk users reactivated by the health-scoring agent

$500K

Costco account reactivated at BP3 Global

$4M

annual revenue at an 80%+ win rate at FlexRadio

13

projects built and documented

01 / Projects

AI agents.
Real work.

13 projects built and documented
GitHub

02 / Method

Discover. Prototype.
Ship. Operate.

  1. 01

    Discover

    Sit with the people doing the work. Find the step that wastes their time.

  2. 02

    Prototype

    Build a working version on the spot, so the conversation is about something real.

  3. 03

    Ship

    Direct AI coding agents through implementation, test the whole path, deploy.

  4. 04

    Operate

    Monitor, troubleshoot and keep improving it after launch.

AI coding agents write the code. I own the solution design, the guardrails, testing, deployment and the business outcome.

03 / Experience

Enterprise sales.
AI solutions.

JoGo.Team

Two-sided marketplace where players book pickup games and small-business venues host them.

AI Solutions Engineer (AI Builder)

2025 - Present
  • Built and operate a production MCP server that lets ChatGPT, Claude and other AI clients work with live platform tools.
  • Created agentic workflows across HubSpot, Stripe, Gmail, SendGrid, WhatsApp, SMS, Google Analytics and Search Console.
  • Real-time customer health and re-engagement system that reactivated 80%+ of at-risk users.
  • Production troubleshooting end to end: reproduce, direct AI coding agents, validate, deploy, monitor.

Customer Success Manager

2024 - 2025
  • Sole customer success function for 1,000+ users across the US, Europe, MENA and Latin America.
  • Onboarding, adoption, engagement, retention and reactivation for venue hosts.
  • Turned customer feedback into product requirements for onboarding, scheduling, payments and engagement.
  • Supported international growth with Spanish and Brazilian Portuguese localization.

RSV AI Consulting

My AI consulting practice.

AI Solutions Consultant

2026 - Present
  • Built and sell an AI BDR that finds accounts with dated buying signals and verifies decision-makers against live sources. Paying clients.
  • Configure bilingual AI receptionists that turn calls and missed calls into leads and appointments.
  • Internal sales and marketing platform: CRM, approved campaigns and deal desk on Cloudflare and Supabase.

BP3 Global

Senior Enterprise Sales Executive, AI & Automation

2024
  • Reactivated a dormant Costco account, led the negotiation and closed a $500K engagement in six months.
  • Advised Fortune 500 customers on automation with UiPath, Automation Anywhere, Pega, Camunda, ABBYY and OutSystems.
  • Ran customer workshops and roadmap reviews with executive and operational stakeholders.

Konica Minolta Business Solutions

Enterprise Sales Executive

2023 - 2024
  • $700K quota across 100+ public-sector accounts: school districts, government and transportation.
  • Sold AI video security, access control, biometric authentication and intelligent document processing.
  • Ran QBRs in person with school district superintendents and leaders.

Quest.ai

Enterprise Sales Executive

2022 - 2023
  • Full enterprise SaaS cycle for an AI Figma-to-React design-to-code platform.
  • Demoed React and Next.js code generation and GitHub integration to engineering buyers.

FlexRadio Systems

Sales Manager, Technical Accounts

2020 - 2022
  • Up to $4M in annual revenue at an 80%+ win rate on software-defined radio systems.
  • Defense, utilities and emergency response customers. Pre- and post-sale technical work: networking, integration, deployment.
  • Presented the product line at the QSO Today Expo 2021.

DistantJob

Enterprise Account Executive

2019 - 2020
  • $1M quota, $150K average deal, placing remote engineering teams with Fortune 500 companies.

3dar

Enterprise Account Executive

2015 - 2019
  • $1M annual quota at 150% of target with Fortune 100 brands.
  • Content, 3D animation, live action and post-production for advertising and film.
  • Led two BDRs and drove 30% annual sales growth.

04 / Principles

Human judgment.
Show the source.

01

A draft is not permission to publish. Every system I build has a human decision point before anything reaches a customer.

02

A researched account is not a booked meeting. Pipeline stages stay honest.

03

Agents do the grunt work so people get their time back for customers.

04

Show the source. Every lead, every number, every claim carries where it came from.

05

A score on a dashboard is not the outcome. The follow-up, and whether the customer came back, is.

06 / Toolbox

Toolbox.

AI and agents

Claude Code · Codex · Gemini CLI · Cursor · Model Context Protocol · Multi-agent orchestration · LLM APIs

Platforms and APIs

Cloudflare · Supabase · Render · Stripe · Twilio · SendGrid · Gmail API · GA4 · Search Console · Sentry · GitHub Actions

Sales and CS stack

Salesforce · HubSpot · Pipedrive · Outreach · Klaviyo

Automation platforms sold

UiPath · Automation Anywhere · Pega · Camunda · ABBYY · OutSystems

Credentials

  • FCC Amateur Extra license, KG5ZZB
  • Electronics Engineering studies, UTN
  • Electronics Technician

Languages

  • English, native
  • Spanish, native
  • Portuguese, working
  • Italian, elementary
RSV AI Consulting

Buying-Signal Research and Lead Qualification

AI BDR

Paying clients

I built and sell an AI BDR around three questions: Is this account a fit? Why would it buy now? Who should sales speak with, and what evidence supports that recommendation?

The problem is not a shortage of names. Sales teams need relevant accounts, credible buying signals and accurate contacts without adding unreliable records or conflicting activity to their workflow.

Qualification comes first. Each client has defined industries, geographies, target roles, qualification gates and suppression rules. Parallel research agents find accounts with dated buying signals, then verify decision-makers against live sources. Source links and verification dates stay with the records; uncertain contacts remain flagged rather than presented as verified.

Research is orchestrated in Claude Code sessions outside the dashboard request path. I designed the platform around a Cloudflare Pages frontend and Supabase Auth/Postgres, with tenant-level Row-Level Security. Client-specific engines connect research, verification, the dashboard and feedback. Clients review delivered accounts and request improvements that inform the next research cycle. Outreach remains human-led in the client's sales stack.

CRM hygiene is part of the design: suppression rules, traceable records and retry-safe updates are safeguards against duplicate or conflicting activity. A researched account is not a booked meeting, a sales-qualified lead (SQL) or revenue. Those stages stay distinct; more names alone do not demonstrate better pipeline.

My AI solutions engineering work covers discovery, qualification logic, workflow and data design, directing AI coding agents through implementation, testing, delivery and ongoing refinement. The product has paying clients.

RSV AI Consulting

CRM, Campaigns and Deal Desk

Sales and Marketing Operations Platform

What has happened with this account? Which outreach is approved? What needs to happen before a proposal becomes a signed agreement? I built an internal sales and marketing platform to connect those decisions in one workflow.

The CRM links accounts, contacts, opportunities and activity. Gmail integration brings email history and open/click tracking into the account context, so engagement can be reviewed alongside the deal rather than in a separate inbox.

The campaign workspace prepares personalized, multi-step drafts from selected audiences. It excludes opted-out audience members during draft generation. Regeneration changes draft items without changing approved or sent records, preserving the boundary between preparation and execution.

The deal desk connects opportunities to quotes, statements of work and electronic-signature records. Sales documents are handled separately from bulk marketing campaigns.

The frontend is a lightweight HTML/CSS/JavaScript SPA on Cloudflare Pages, using the Supabase SDK. Supabase Auth issues user sessions; Postgres Row-Level Security scopes tenant data. Cloudflare Pages Functions handle server-side integrations, including Gmail OAuth. Tenant-isolation tests check customer-data separation. Architecture decision records document choices such as separating sales documents from marketing. This is the working platform supporting my consulting operations alongside the client-facing BDR product.

My AI solutions engineering contribution spans workflow discovery, data modeling, integration boundaries and interface design. I direct AI coding agents through implementation, define and review tests, and handle troubleshooting and refinement.

RSV AI Consulting

Lead Capture and Appointment Booking

Bilingual AI Receptionist

What happens when a prospect calls while staff are busy? Can the business capture the inquiry, understand the request and move it toward an appointment in the caller's language? I sell and configure a bilingual AI receptionist around that workflow.

The solution supports English and Spanish and combines AI voice with missed-call text-back, voicemail transcription and call summaries. Calendar integration and appointment reminders connect the conversation to the booking process. Multi-location configurations support routing and separate calendars.

The implementation starts with discovery: which calls the business receives, what information needs to be captured, how appointments are booked and when a person should take over. Those requirements become call flows, lead-capture fields, booking rules and handoff procedures. The design maps voice and missed-call entry points to contact information, routing and calendar actions; testing follows the complete inquiry-to-handoff path.

My contribution is AI solutions engineering and customer implementation: demonstrations, scoping, proposals, platform configuration, integrations, testing, troubleshooting and onboarding. I build the customer workflow on established voice and telephony platforms.

The practical output is a configured path from an incoming or missed call to a usable lead, appointment or human handoff, with the call context available for follow-up.

JoGo.Team

Health Scoring, SMS and WhatsApp

AI Customer Reactivation

80%+ of at-risk users reactivated

Which users need attention before they disengage? What signal should trigger follow-up, and how do we know whether they returned? I designed and operate a customer-health and reactivation system for JoGo.Team to connect those questions to action.

The workflow evaluates activity and engagement, identifies at-risk users and connects the health assessment to automated SMS and WhatsApp follow-up through Twilio. It tracks whether users return and surfaces cases needing individual attention for human follow-up.

The workflow runs within JoGo.Team's Django application on Render, with Twilio providing SMS and WhatsApp delivery. Its architecture connects platform activity, health-scoring logic, messaging and the re-engagement outcome. The useful output is a timely follow-up workflow, not simply a score on a dashboard.

More than 80% of the users identified as at risk were reactivated. This is the reactivation result for that identified group, not a company-wide retention or revenue figure.

My AI solutions engineering work covers the scoring and messaging requirements, workflow design, directing AI coding agents through implementation, end-to-end testing and deployment. I continue to monitor and maintain the production system.

The project applies customer-success judgment to lifecycle marketing automation: identify where attention is needed, act through the appropriate channel and check the return behavior.

JoGo.Team

Google Search Console and GA4

Agentic SEO and Growth Analytics

Did search visibility change, did website activity change, or are we looking at a different mix of queries and pages? I built and operate a search and growth-analytics workflow for JoGo.Team that turns those questions into repeatable investigations.

Agents retrieve first-party data through Google Search Console and Google Analytics 4 APIs, compare reporting periods and surface changes for review. Query-level and page-level analysis helps narrow an investigation beyond the headline traffic total.

The workflow includes sitemap checks, indexing investigations, keyword overlap and cross-checks between search data and site activity. I use purpose-built analyses to test explanations and identify the next technical or content action.

The architecture separates Search Console/GA4 API retrieval, period comparisons, focused audit scripts and saved JSON reporting outputs. Collection and analysis are distinct steps, so findings can be checked against the retrieved data. Search visibility and on-site behavior remain distinct measures; interpreting the difference is part of the work.

My AI solutions engineering contribution is defining the business question, selecting checks, directing AI coding agents through implementation and validating the output before recommending an action. I maintain the workflow as part of ongoing growth operations.

The deliverable is an evidence-backed investigation and a clear next step, rather than a claim that a traffic movement proves a particular cause.

JoGo.Team

Customer and Operational Workflows

Production MCP Integration

Live in production

How can an AI assistant work with a live platform, and where must a person remain in control? I built and operate a production Model Context Protocol (MCP) server connecting compatible AI clients, including ChatGPT and Claude, to JoGo.Team tools and workflows.

The server exposes defined platform capabilities through a tool interface. This connects an assistant's conversation to operational workflows instead of stopping at a written recommendation.

A draft-then-publish gate separates preparing material from publishing it. That approval boundary is part of the workflow design, so preparation and customer-facing execution are treated as different steps.

The MCP endpoint is implemented in Django on Render behind Cloudflare, using JSON-RPC tool dispatch. Public read tools are separate from authenticated draft creation. The implementation includes scoped, revocable host credentials stored as hashes, rate limits and call logging. The assistant creates a draft using the same form validation as the website, keeping required fields and timezone/currency rules consistent. The host finishes publishing in the web interface; untrusted text reaching an assistant does not become permission to publish. LLM reasoning stays in the connected client, while the server handles tools and application rules.

My AI solutions engineering contribution covers requirements, tool and API integration, directing AI coding agents through implementation, testing and deployment. I also monitor the production integration and troubleshoot failures.

The project demonstrates a working connection between AI assistants and customer operations, with an explicit human decision point for publishing.

GitHub

Customer Operations, Integrations and Growth

AI Workflow Portfolio

What can someone inspect beyond a project description? This public portfolio brings together AI-assisted implementations and technical demonstrations covering customer operations, lifecycle automation, API integrations and external-data workflows.

The examples make the work concrete: how an operational question becomes a workflow, how systems exchange information and where a reviewer can inspect the implementation and its documented boundaries. Separate repositories let a technical reviewer examine each example's setup and integration approach in context.

Runnable examples use synthetic data. Their documentation distinguishes the demonstration's scope from the production systems described in the individual project entries on this profile, keeping customer information out of public examples.

The collection includes customer-account intelligence, messaging-based re-engagement and integration case studies. It is a set of inspectable examples, not a claim that every repository is a separate deployed commercial product.

My AI solutions engineering contribution spans problem definition, workflow design, directing AI coding agents through implementation, testing and troubleshooting. The portfolio complements the business context and outcomes documented in the individual case studies.

Explore the repositories and documentation: https://github.com/jeypnet

Internal

Editorial Review and Publishing Workflow

AI Content Operations

How can a team keep producing content while preserving its voice, checking claims and deciding what actually gets published? I built an AI-assisted content-operations workflow with editorial approval at its center.

The system prepares LinkedIn copy and matching visuals for separate brand voices, then queues the drafts in a review dashboard. A reviewer can edit, approve or reject each item before the publishing workflow selects approved material. Publication status and links are recorded for follow-up.

Headless Claude sessions prepare drafts; a SQLite queue tracks draft, approved, posted, rejected and failed states. A local review dashboard handles editorial decisions, and the browser-based publishing step selects approved items. Scheduled jobs separate generation from posting. Topic rotation limits repetition; a curated facts bank constrains numerical claims to approved information.

The key control is the handoff: producing a draft is not permission to publish it. Human review covers the message and visual before the workflow proceeds.

My AI solutions engineering work covers the editorial requirements, workflow design, review controls and integration between generation and publication. I direct AI coding agents through implementation and test the handoffs.

The result is a governed content-production process that combines repeatable preparation with human editorial judgment.

JoGo.Team

Vertical Reels for Social Marketing

AI-Assisted Video Production

Which moments in raw footage are worth using, how should they fit a vertical screen, and what makes the final edit ready to publish? I built a local video-production pipeline for JoGo.Team social marketing around those decisions.

Audio and motion signals identify candidate moments. Contact sheets make the candidates easier to review before selected clips are assembled with captions, an opening hook and a call to action.

ffmpeg handles ingestion and audio/motion detection. Contact sheets batch candidate frames for Claude's visual review, which produces a JSON edit manifest defining segments, captions, hook and call to action. CUDA/YOLO-assisted reframing and NVENC encoding run locally on the GPU, producing 1080 x 1920 H.264 video. AI review and rendering are separate stages, with per-stage cost tracking.

Creative selection remains a review step. A detected moment is a candidate, not a finished story; the assembled output is checked before publication.

My AI solutions engineering contribution covers the production requirements, workflow design and directing AI coding agents through implementation. I incorporated review points and cost visibility so the process can be evaluated beyond whether it produces a video file.

The output is a vertical reel prepared for social marketing, connecting raw footage to a reviewed creative asset.

Product

Calendar Integration and Booking Workflows

Appointment Scheduling

Live

Is that meeting time actually available in both people's time zones, and what happens when someone reschedules? I built a scheduling application that connects a public booking page to Google Calendar availability.

The availability engine combines working hours with busy periods and accounts for daylight-saving transitions, buffers and booking limits. The booking workflow creates calendar events and meeting links, then supports confirmation, rescheduling and cancellation.

Cloudflare Pages and Pages Functions serve the application; Supabase Postgres stores scheduling data. Google Calendar OAuth2 authorizes calendar access, and the freeBusy API supplies busy periods. A separate Cloudflare Worker performs background synchronization. Calendar event creation and reschedule/cancel workflows connect availability calculations to the actual booking.

Slot calculations have dedicated tests for daylight-saving changes, buffers and limits. These rules are central to the product: a booking page is useful only when its offered times reflect the scheduling constraints.

My solutions engineering work covers booking requirements, calendar integration, workflow design and directing AI coding agents through implementation. I define and review the slot-engine tests and validate the booking flow.

The application supports sales demos and customer appointments by turning booking intent into a confirmed calendar event with a clear path to change or cancel it.

Operations

Catalogs, Listings and Reporting

Multi-Channel Commerce Workflows

Is the product data ready for the marketplace, what needs attention in the seller account, and what changed since the last review? I built and used commerce tools and workflows to make those operational questions easier to answer.

The work covers product catalogs, marketplace listings, account audits and recurring performance reporting, with integrations and tooling around eBay, Printify and Amazon seller workflows.

The technical work includes Amazon Selling Partner API, eBay API and Printify-related tooling. Focused scripts handle catalog/product data, account and listing audits, backups and recurring reporting. Marketplace-specific requirements stay separate from shared operational checks, so each channel's constraints are handled in context.

Listing and account checks help identify the next action. I review outputs before making marketplace changes, keeping the distinction between an audit finding and an approved update.

My solutions engineering contribution covers operating requirements, product-data handling, integration workflow design and directing AI coding agents through implementation. I validate the outputs and use recurring checks to support ongoing commerce operations.

This project represents practical multi-channel commerce tooling, including print-on-demand operations, rather than an unverified claim of a fully autonomous store.

RSV AI Consulting

Technical SEO and Machine-Readable Content

AI Search Readiness

Can an automated system understand what a website offers and find its important content? I implemented machine-readable discovery assets for an AI consulting website to make that information more explicit.

The work includes llms.txt, an AI-oriented sitemap and a structured SEO manifest alongside conventional robots and sitemap files. These assets describe the site's products and content in formats intended for automated interpretation.

The discovery layer consists of published text and JSON assets alongside the public website: llms.txt, ai-sitemap.json and seo_manifest.json. It complements robots and conventional sitemap files rather than introducing a separate conversational application. Content structure and published-file validation are part of the implementation, not just the presence of a filename.

My solutions engineering contribution covers deciding what information to expose, structuring the content, directing AI coding agents through implementation and checking the published output.

This is a technical SEO and AI-discovery implementation project. Its verified output is the published, machine-readable content; adoption by individual AI systems and discovery performance are separate questions to evaluate.

The work complements first-party search analytics by improving how the website describes and organizes its information for automated consumers.

Sales tool

Excel-Based CPQ and Quote Preparation

AI Quoting Assistant

I built an AI-assisted CPQ (configure, price, quote) tool that turns a large Excel source of truth into a simple sales-facing interface for preparing quotations.

The problem: the information needed for a quote lives in a massive workbook, but the salesperson needs a practical way to use it. How do we keep the authoritative data while making quotation preparation faster and easier?

The architecture separates the reference data from the user experience. Excel supplies the source of truth; the frontend makes that information usable in the quote-preparation workflow. Salespeople work through the interface rather than navigating the full spreadsheet for every quotation.

The central design requirement is continuity with the source: simplifying the experience should preserve the workbook's role as the authoritative reference.

My AI solutions engineering contribution is translating that sales-process bottleneck into a working tool, defining the source-to-interface workflow and building the frontend experience around the people preparing quotes.

The tool is designed to accelerate quotation preparation and make a complex spreadsheet-backed process easier to use.

Agent console

Juan Pablo Pacheco

AI Solutions Engineer