Things I've built to solve actual problems.

Each one starts with the problem, not the tool. The stack follows.

Full StackAnalyticsIntegrations

Problem: Recruitment performance and operational information existed across several systems.
What I built: A custom analytics application that consolidates recruitment reporting and performance visibility.
Result: Reporting moved from weekly manual exports to a live application.
Stack: Next.js, React, APIs, RecruitCRM, PostgreSQL, Custom Analytics

Pipeline by stage

WeekMonthQuarter

Sample numbers. The real one runs on private client data.

AIAutomation

Problem: Difficult handwritten information required manual interpretation.
What I built: An AI-powered OCR workflow that extracts structured information.
Result: Roughly 90% of fields on the tested document set were extracted correctly with no human involvement.
Stack: n8n, OCR, AI / LLM, APIs, Structured Outputs, Workflow Orchestration

ScanOCRLLM ExtractionValidationReview QueueDatabase

GTMAutomationIntegrations

Problem: Lead generation ran as a chain of disconnected manual steps: find the leads, enrich them, verify emails, upload to the CRM, add to sequences, follow up, report.
What I built: An end-to-end lead pipeline built around n8n and a small Node.js service: sourcing tools feed enrichment, verification filters out what would bounce, clean records land in the CRM, and outreach sequences start automatically with follow-ups and reporting attached.
Result: Leads flow from source to sequenced outreach without manual handoffs, and the team works the conversations instead of maintaining the list.
Stack: n8n, Node.js, APIs, RecruitCRM, Smartlead, HeyReach, SalesQL, NeverBounce, Evaboot

Lead SourceEnrichmentVerificationCRMOutreachFollow-UpAnalytics

IntegrationsAutomationAnalytics

Problem: Inbound forms triggered text messages, but nobody could tell which messages led to bookings, who clicked and didn't book, or where people dropped off.
What I built: Form submissions trigger a Make workflow that sends SMS through Twilio using tracked links.
Result: Every message, click, and booking is attributed, and follow-ups go out on schedule instead of when someone remembers.
Stack: Twilio, Make, Calendly, Node.js, Custom Backend, Tracking Dashboard

FormWorkflowTwilioTracked LinkCalendlyAnalyticsFollow-Up

AnalyticsAutomationIntegrations

Problem: Campaign data lived in Meta Ads, Google Ads, and Shopify, each with its own definition of a conversion.
What I built: Automated campaign reporting: scheduled pulls from each platform's API, a normalization step for currencies, time zones, and attribution windows, storage in a database, and reports generated from it.
Result: Reports generate themselves on schedule, and ad spend finally sits next to the revenue it produced.
Stack: Meta Ads, Google Ads, Shopify, APIs, Scheduled Workflows, Custom Reporting

Ad PlatformsScheduled PullNormalizeDatabaseReport

AutomationIntegrations

Problem: A critical system had no API.
What I built: Python RPA scripts drive the portal where no API exists.
Result: The daily manual session became a scheduled job with error alerts, and the person who used to do it no longer needs to.
Stack: Python, n8n, APIs, RPA, HTTP Requests

ScheduleRPAExtractAPI CallsDestinationLog