Jordi – Workflow Automation, SQL, Webhooks

Print-on-demand platform connecting sellers with producers, design tools, and marketing channels. Delivered AI solutions for visual content generation, internal knowledge retrieval, and fulfillment workflow automation to improve scalability and operational efficiency.

• Built an automated AI image generation pipeline to produce custom, copyright-free product visuals at scale: started with GPT Image 1, then migrated to Stable Diffusion via Replicate API to reduce per-image generation costs while maintaining output quality, eliminating reliance on stock image libraries entirely;
• Implemented a RAG system over internal SOPs and product documentation using OpenAI API, LangChain, and FAISS, reducing internal query resolution time by approximately 40%;
• Built an n8n-based fulfillment automation pipeline eliminating manual order routing between suppliers and clients, cutting processing overhead by roughly 35%;
• Integrated Claude AI API for a context-aware customer-facing chat interface, enabling natural language product queries mapped to live inventory;
• Designed and led AI architecture across 3 core product areas: discovery, design tooling integration, and marketing platform automation;
• Structured the AI product roadmap for a fast-moving startup environment, balancing speed with production-grade reliability.

Consulting practice delivering AI and automation solutions to startups and growing businesses across North America, Europe, and West Africa, serving 10+ clients across e-commerce, operations, marketing, and customer support and contributing to over $250K in combined client revenue.

• Deployed Make.com and OpenAI integration for an asset management firm, eliminating roughly 25 hours per week of manual data-routing and reducing operational overhead by about 60%;
• Built RAG knowledge systems (OpenAI API, Pinecone, FastAPI) enabling non-technical staff to query internal SOPs, contracts, and compliance documents in natural language;
• Architected autonomous n8n agents for end-to-end lead qualification, CRM updates, and follow-up sequences, with zero manual hand-off from form submission to qualified pipeline entry;
• Deployed Voice AI agents (Voiceflow + ElevenLabs) for SMB clients, reducing inbound support volume by 60% and cutting first-response time from hours to seconds;
• Developed Python lead-scoring microservices using scikit-learn, improving outreach conversion by roughly 40% through behavioral and demographic signal processing;
• Scaled AI-integrated e-commerce operations to multi-six-figure revenue using ML demand forecasting (Prophet), automated ad targeting pipelines, and inventory alerting;
• Managed non-technical client relationships across 3 continents, translating complex automation architecture into plain business language for every stakeholder.

End-to-end AI prospecting and CRM automation system for a Canadian network marketing company, automating lead enrichment, AI-powered outreach, meeting scheduling, and follow-up workflows, with human involvement limited to the final presentation call.

• Designed the full prospecting pipeline: form submission triggered a Make.com webhook, contact data was pushed to HubSpot CRM, and a Vapi AI voice agent call was initiated automatically;
• Configured and deployed the Vapi voice agent: custom voice persona scripted to introduce the business opportunity, qualify the prospect, and pitch a presentation call booking in a single outbound call;
• Integrated Calendly API with Vapi so the agent offered available time slots in real time during the call and confirmed bookings directly, with no human scheduling involved;
• Built the HubSpot automation layer via Make.com: every prospect interaction (call completed, declined, booked, no-show) automatically updated the contact record, triggered the appropriate follow-up sequence, and moved the deal stage;
• Implemented the post-call data pipeline: Vapi transcript passed through Make.com, summarized by OpenAI API, and logged as a structured note on the HubSpot contact with disposition tag and next action;
• Built re-engagement automation for non-bookers: 24-hour delay triggered an automated WhatsApp message with a direct Calendly link;
• Designed error handling and retry logic: failed calls flagged in HubSpot with a retry tag and re-queued for the next call batch, with no leads dropped silently.

Centralized Slack-based operations platform that unified workflows across Slack, WhatsApp, Google Workspace, Asana, and CRM systems, enabling teams to receive updates, make decisions, and trigger actions from a single interface.

• Designed the full hub architecture: Make.com and n8n as the automation backbone, routing signals from HubSpot, Stripe, Airtable, and Supabase into structured Slack Block Kit notifications with interactive buttons for one-click actions;
• Built the revenue signal layer: Stripe webhook events (payment received, failed, disputed, subscription churned) routed to the appropriate Slack channel with full context and immediate response actions available without opening Stripe;
• Integrated Claude AI API as a natural language command interface within Slack: team members typed plain-language requests and Claude interpreted the intent and triggered the corresponding Make.com scenario or n8n workflow;
• Implemented role-based access controls: Slack channel permissions and action scopes defined by role, ensuring each team function only saw and triggered what was relevant to them;
• Built the failure alert system: any broken webhook, failed automation step, or data sync issue surfaced immediately in a dedicated Slack channel with full context, a retry action, and an escalation path;
• Built the Airtable-to-Supabase migration pipeline as part of the engagement, managed entirely through Make.com webhooks.

AI-powered content operations platform automating the full content lifecycle, from ideation and scripting to generation, repurposing, distribution, and team workflow management, with automated Slack, WhatsApp, and email notifications enabling a lean content team to operate without technical expertise or dedicated project management.

• Designed multi-agent architecture: separate agents for ideation (trend analysis), scripting (Claude AI API), short-form repurposing (GPT-4), and distribution scheduling (n8n + platform APIs);
• Built the trend intelligence module: automated ingestion of YouTube trending data, Google Trends, and Reddit signals, summarized by an LLM into a weekly content brief delivered to the team automatically;
• Implemented the repurposing pipeline: long-form video transcript via Whisper API automatically reformatted into Twitter threads, LinkedIn posts, email newsletters, and YouTube Shorts scripts;
• Built the team notification and handoff system: when a video was ready for editing, simultaneous notifications fired across Slack, WhatsApp, and email; the team configured which channels were active per role to control costs;
• Engineered the full pipeline handoff logic: editor marked work complete and an automated message fired to the reviewer; reviewer submitted feedback and it routed back to the editor via Slack with full context; the same logic applied for every production role in the pipeline;
• Used Airtable as the central production tracker: every content piece, status update, assignment, and feedback loop logged and visible to the whole team in one view;
• Integrated ElevenLabs voice cloning for AI-narrated short-form audio, maintaining creator voice consistency across formats.

End-to-end AI sales and client management system for an asset management company, featuring an AI WhatsApp sales agent, CRM automation in Airtable, and operational AI workflows that automated lead qualification, client onboarding, customer support, and sales activities.

• Architected the full conversational AI system: ManyChat managing the WhatsApp channel and flow routing, OpenAI API (GPT-4o mini) powering the reasoning layer for dynamic, context-aware responses;
• Built the system's knowledge base: products, services, pricing, FAQs, and objection-handling scripts structured as context fed into the OpenAI prompt chain;
• Designed the autonomous sales flow: prospect sends first message, system qualifies with targeted questions, presents the relevant offer, handles objections, and closes or schedules a follow-up with no human involvement;
• Implemented two-tier human handoff logic: customer requests a human triggers a warm transfer with full conversation context; confidence drop or out-of-scope query triggers automatic escalation to a human agent;
• Built the Make.com orchestration layer: every conversation event extracted structured data and updated the Airtable CRM record in real time: lead stage, summary, contact info, and outcome logged automatically;
• Built and maintained additional AI-powered automations across the business: data enrichment pipelines, automated client reporting sequences, and internal operational workflows;
• Maintained, iterated, and extended the system throughout the engagement: added new product knowledge, refined prompt logic based on real conversation data, and deployed new automation flows as the business evolved.

24/7 AI booking and support agent deployed on WhatsApp, automating appointment scheduling, customer support, follow-up reminders, and post-service engagement through integrated CRM and workflow automation systems.

• Deployed WhatsApp Business API and ManyChat as the channel layer: WhatsApp Business API handled message delivery and template management, while ManyChat managed conversation flow routing, session state, and the handoff conditions between AI and human agents;
• Integrated OpenAI API (GPT-4o) as the reasoning engine, trained on the business service catalogue, pricing, booking policies, and common objections, with prompt chaining to maintain conversation context across multi-turn interactions;
• Integrated Calendly API for real-time availability checks and booking confirmations directly within the WhatsApp conversation, enabling a prospect to go from first message to confirmed appointment in under 60 seconds;
• Built the Make.com orchestration pipeline: confirmed booking triggered HubSpot contact creation, a confirmation WhatsApp message, a 24-hour reminder, and a post-appointment follow-up automatically;
• Implemented human handoff logic: customer requests a human, agent detects a sensitive or ambiguous situation, or confidence drops below threshold all triggered a warm escalation to a staff member with the full conversation transcript delivered via WhatsApp notification;
• Designed data scoping: client PII stored in HubSpot only, never retained in ManyChat or Make.com data stores, giving the business a clean and auditable data trail;
• Built the fail-safe layer: if the booking API failed mid-conversation, the agent acknowledged gracefully, logged the issue to Airtable, and triggered a human callback task in HubSpot so no customer was left without a resolution path.

Cargo management platform for logistics operations, enabling real-time shipment tracking and automated WhatsApp status updates, with three platform implementations delivered across Airtable, Hostinger Horizons, and Base44 based on client requirements and budget.

• Designed and built v1 on Airtable for the primary client: customer records, package intake forms, shipment tracking, and agent dashboards, all connected to a Make.com automation backbone;
• Shipped v2 for a second client on Hostinger Horizons and v3 for a third client on Base44, adapting the same business logic and automation architecture to different no-code stacks based on each client's context and budget;
• Built the WhatsApp Business API notification pipeline via Make.com: every new package intake triggered an instant WhatsApp message to the customer containing tracking number, package details, and estimated delivery;
• Integrated OpenAI API to generate personalized message content for every outgoing WhatsApp notification: pulled client data from Airtable records and generated a custom message body for each interaction so no two customers received the same generic text;
• Built a custom order tracking interface on the client website, pulling live Airtable shipment data and displaying real-time status to customers without requiring a login;
• Engineered exception and alert automations: delayed shipments, flagged packages, and agent assignment changes triggered internal notifications to keep the ops team ahead of issues.

Growth and operations automation platform for an early-stage SaaS company, automating lead follow-up, customer onboarding, subscription management, and reporting through a connected, self-maintaining workflow ecosystem.

• Built the inbound lead pipeline: form submission triggered HubSpot contact creation, OpenAI generated a personalized first-touch email based on the prospect's stated use case, and the contact was enrolled in the appropriate nurture sequence, all within 90 seconds of form submission;
• Implemented Stripe and HubSpot sync via Make.com: subscription events automatically updated the deal record, recalculated MRR, and triggered the correct customer success action;
• Built a weekly business health report: scheduled n8n workflow queried Supabase for key metrics, fed the data to OpenAI for plain-language commentary, formatted the output, and delivered it to the founding team via Slack and email every Monday morning;
• Designed the customer onboarding automation: new subscriber triggered a multi-step sequence across email and WhatsApp, with completion checkpoints tracked in Airtable and stalled onboardings flagged to the CS team via Slack;
• Built full error handling across every workflow: failed steps surfaced in a dedicated Slack alert channel with context and a manual retry option, with no revenue-critical process failing silently.

AI tool discovery platform enabling users to describe goals in natural language and receive ranked, contextually relevant tool recommendations through automated categorization, ingestion, and semantic search across thousands of listings.

• Architected the full AI discovery engine: tool listings ingested via automated scraping pipeline (Python + Playwright), embedded using OpenAI Embeddings API, stored and queried via Pinecone vector database;
• Built the semantic search layer: natural language query returned top-ranked tool matches with a GPT-4 generated contextual explanation for each result;
• Designed the automated categorization pipeline: new tool submissions classified by use case, pricing model, and technical complexity using a LangChain prompt chain;
• Implemented a Make.com pipeline for weekly automated discovery of newly launched AI tools from ProductHunt, newsletters, and RSS feeds, keeping the knowledge base current without manual curation;
• Built the Next.js and React frontend with a conversational search interface powered by streaming OpenAI responses;
• Deployed FastAPI backend on AWS Lambda with Supabase and PostgreSQL for structured metadata and Pinecone for vector search.

VC-backed B2B marketplace for event service providers across Romania and Eastern Europe, where I built and owned the end-to-end data infrastructure, automating vendor discovery, extraction, and ingestion to populate the platform from zero listings within a 90-day launch timeline.

• Built Python automation workflows that continuously scraped vendor profiles from business directories, trade platforms, and supplier websites across Romania, cleaning and structuring the data for platform ingestion;
• Designed the data normalization layer: raw scraped HTML converted into deduplicated, categorized structured records inserted into PostgreSQL via an automated ETL pipeline;
• Generated 2,000+ structured vendor database entries per day at peak, enabling the marketplace search engine to be fully populated before public launch;
• Implemented error handling, retry logic, and proxy rotation to maintain scraping reliability and consistent throughput across anti-bot countermeasures;
• Delivered the complete data infrastructure pipeline within the 90-day window, enabling the platform to launch with a fully operational vendor catalog.

Blockchain-powered marketplace for tokenized real-world assets, enabling the trading of real estate, luxury goods, fine art, and collectibles through NFT-backed ownership, with AI-driven asset onboarding, valuation, and transaction workflows.

• Built the full-stack platform: React and Next.js frontend, Node.js backend, PostgreSQL database, with real-time price tracking and transaction history via WebSocket feeds;
• Designed the AI-assisted asset onboarding pipeline: computer vision model that verified physical item authenticity from uploaded imagery before NFT minting was triggered;
• Built an automated price intelligence system using real-time market data and an XGBoost regression model for dynamic asset valuations and owner price-movement alerts;
• Architected the smart contract interaction layer (Solidity + ethers.js) connecting the onboarding pipeline to NFT minting, with webhook-based event triggers for ownership status updates;
• Integrated OpenAI API for AI-generated asset descriptions and provenance narratives, reducing listing creation time from roughly 45 minutes to under 3 minutes per item;
• Built n8n orchestration for the full trading lifecycle: bid received, smart contract execution, ownership transfer, and physical custody notification, all automated end-to-end.