I turn manual operations into intelligent systems that run themselves.

Software · SaaS Migration

Trello → Monday.com Migration Platform

The problem

Get Levrg needed to move 70+ client workspaces from Trello to Monday.com. Manual migration meant recreating every board, card, comment, and attachment by hand — weeks of error-prone work per batch, with no audit trail and no way to keep client and internal boards linked.

The solution

I designed and built a Laravel web application on a service-layer architecture with dependency-injected TrelloService, MondayService, DataMapper, and AuditLogService components, driving Trello REST and Monday.com GraphQL APIs. It supports configurable status mapping, task-type synchronization, dual client–internal boards with item linking, fuzzy name matching for member assignment, and rich imports of descriptions, comments, and attachments — deployed to AWS with GitHub Actions CI/CD.

How it works

• Service-layer architecture with dependency-injected Trello/Monday services
• REST + GraphQL API orchestration with board-level caching
• Configurable status mapping and task-type synchronization
• Dual client–internal boards with item linking and fuzzy member matching
• Rich imports: descriptions, comments, and attachments
• Real-time migration progress, job history, and full audit logging
• AWS deployment with GitHub Actions–based CI/CD

Results

• 70+ client workspaces migrated successfully
• Weeks of manual recreation replaced by monitored, auditable jobs
• Client and internal boards stay linked after migration

Stack

AI · Workflow Orchestration

Meeting → Action-Items AI Pipeline

The problem

After every meeting, someone had to re-listen, write notes, extract action items, post them to the right project board, and chase owners. Across 20+ people and daily meetings, post-meeting admin consumed hours and items regularly slipped through.

The solution

I built and deployed an AI pipeline that ingests Fireflies.ai transcripts, uses Claude and OpenAI to generate notes, summaries, and action items, routes each action item to the correct Monday.com task board by meeting name, assigns it to the right owner, and delivers summaries via Notion, Gmail, and Slack.

How it works

• Fireflies.ai transcript ingestion for every recorded meeting
• LLM processing (Claude + OpenAI) for notes, summaries, and action items
• Automatic board routing by meeting name
• Owner detection and per-item assignment
• Delivery to Notion, Monday.com, Gmail, and Slack

Results

• Rolled out to 20+ internal users across teams and CXOs
• Post-meeting admin and duplicate data entry sharply reduced
• Action items land on the right board, assigned, before the room empties

Stack

Read the full case study

AI · Sales Automation

AI Follow-Up Agent for Lead Nurturing

The problem

Leads in GoHighLevel went cold because follow-up depended on humans remembering to call, text, and email — and generic drip campaigns ignored what each lead had actually done.

The solution

I developed an AI Follow-Up Agent using n8n that watches lead interactions in GoHighLevel and generates contextual follow-up with OpenAI — placing voice calls with ElevenLabs speech through Twilio, and sending personalized SMS and email sequences that adapt to each lead's behaviour.

How it works

• Lead-interaction triggers from GoHighLevel
• OpenAI-generated contextual messaging per lead
• AI voice calls via Twilio + ElevenLabs
• Multi-channel sequences: call, SMS, and email
• Behaviour-based branching instead of fixed drips

Results

• Every lead gets timely, personalized follow-up across three channels
• Sales team freed from manual sequence management
• Follow-up quality no longer depends on rep workload

Stack

Read the full case study

Automation · n8n

Monday.com Board Sync Suite

The problem

Client-facing and internal Monday.com boards drifted apart constantly: statuses updated on one side but not the other, renamed or deleted items left orphans, and project boards never reflected the true state of task boards.

The solution

I designed an end-to-end n8n automation suite on Monday.com webhooks: client update propagation with user mentions, bi-directional status synchronization via configurable mappings, dynamic item movement across groups, cascading renames and deletions for linked items, plus a second scalable automation aligning Project and Task boards with dynamic column/label management and scheduled syncs (daily and 6-hour intervals). Failures alert to Slack in real time.

How it works

• Webhook-driven item creation, update, and deletion handling
• Bi-directional status sync via configurable mappings
• Cascading renames and deletions for linked items
• Dynamic project column and status label management
• Scheduled syncs (daily + 6-hour) for portfolio-to-task alignment
• Real-time failure alerts to Slack

Results

• Cross-board data consistency without manual reconciliation
• Project health tracking reflects reality, on schedule
• Failures surface in Slack immediately instead of silently drifting

Stack

AI · Research Automation

STR Compliance Research Automation

The problem

Researching short-term-rental (STR) compliance rules per jurisdiction meant hours of manual searching, reading, and report writing — repeated for every new market, with inconsistent quality and no citations.

The solution

I automated the whole workflow in n8n: the Perplexity API performs the research, results are structured into professional, source-cited reports written straight into Google Docs, and Slack notifies the team the moment a report is ready.

How it works

• Perplexity API research with source citations preserved
• Structured, professional report generation into Google Docs
• Slack notifications on completion
• Repeatable per-jurisdiction runs with consistent quality

Results

• Manual research time reduced dramatically — hours to minutes
• Every report is source-cited and consistently formatted
• New jurisdictions covered on demand

Stack

Automation · Data Extraction

Spotify Playlist Curator Intelligence Tool

The problem

A music-marketing team was manually hunting playlist curators — 40+ hours per week of searching, cross-checking, and copying contact data, and the pipeline of verified curators still ran dry.

The solution

I built a Python automation tool combining Playwright browser automation with the Spotify API to discover playlists, extract and verify curator information, and compile everything into a clean, deduplicated dataset ready for outreach.

How it works

• Playwright-driven discovery across playlist ecosystems
• Spotify API enrichment and verification of curator data
• Deduplication and validation for outreach-ready records
• Re-runnable pipeline to keep the dataset fresh

Results

• 3,000+ verified playlist curators compiled
• Manual research cut from 40+ hours per week to minutes
• Outreach efficiency and playlist placement rates improved

Stack

RPA · Data Pipeline

Utility Bill Data Pipelines (PG&E + SFWater)

The problem

The client needed billing data from PG&E and SFWater for a large set of accounts. Staff were logging in to two portals manually, downloading PDF bills one by one, re-typing values into a database, and filing documents — slow, error-prone, and impossible to scale.

The solution

I built unattended Robomotion bots for both providers: each logs in to its portal, iterates through every account from an Excel-driven config, downloads PDF bills, runs Tesseract OCR to extract billing values, validates and inserts data into PostgreSQL, and archives source PDFs to Google Drive and DigitalOcean Spaces. Discord notifications report every start, resume, and error, and both providers feed one shared schema for unified reporting.

How it works

• Automated portal login and account iteration for two utility providers
• PDF download and layout-aware OCR extraction with Tesseract
• Field parsing, validation, and inserts into a shared PostgreSQL schema
• Dual archival to Google Drive and DigitalOcean Spaces
• Discord alerting with automatic resume after interruptions

Results

• Bill retrieval and data entry became fully hands-off for both providers
• One clean, queryable database across PG&E and SFWater
• Errors surface in Discord within seconds instead of weeks later

Stack

Read the full case study

RPA · Property Management

Automated Bill Submission on AppFolio

The problem

Real-estate administrators were submitting vendor bills on AppFolio.com by hand: log in, open the bill form, type vendor and invoice details, attach files, save, repeat. Across multiple properties this consumed hours every week and invited data-entry mistakes.

The solution

I built a Robomotion bot (JavaScript) that logs in with admin credentials, navigates to bill entry, fills vendor and invoice fields, uploads invoice attachments, saves the bill, and immediately starts the next one — with error detection around each submission so a single failure never silently corrupts a batch.

How it works

• Automatic login and navigation to the bill-entry workflow
• Vendor, invoice, and amount fields filled from structured input
• Automated file uploads for invoices and receipts
• Per-bill error detection with safe continue/retry behaviour
• Batch mode: submit bill after bill without interruption

Results

• Bill entry across multiple property portfolios runs as a batch job
• Attachment handling and field entry are consistent on every bill
• Scales to any number of properties without extra headcount

Stack

Software · Industrial IoT

Villicus Industrial Monitoring Server

The problem

An industrial automation client had a legacy C# TCP server bridging client devices and Micro800-series PLCs. It needed to move to Python while preserving robust communication with an existing PHP client — and industrial environments punish flaky networking and unhandled errors.

The solution

I converted the C# project to a multithreaded Python TCP server that maintains persistent connections, reads and writes PLC tag values, and streams tag updates in real time. JSON serialization plus zlib/base64 compression keeps bandwidth low; robust exception handling keeps the server alive through network faults; configuration is externalized for deployment across plant networks.

How it works

• Full C# → Python conversion with a compatible PHP client protocol
• Real-time TCP/IP communication with Micro800 PLCs
• Read, write, and subscribe semantics for PLC tag values
• JSON serialization with zlib + base64 compression
• Multithreaded concurrent client handling with fault tolerance

Results

• Operators monitor and control PLCs remotely and safely
• The server handles many concurrent clients without degrading
• Deployed as a practical, real-world industrial solution

Stack

Software · Healthcare

Clinic Management System

The problem

A clinic was coordinating patients, appointments, and records across disconnected tools and paper processes — slow for staff and opaque for management.

The solution

I designed and implemented a comprehensive Clinic Management System using Laravel, Livewire, and Alpine.js — a reactive, server-driven interface without a heavy SPA framework, covering the clinic's operational workflows end to end with a user-friendly experience.

How it works

• Full-stack TALL-adjacent architecture: Laravel + Livewire + Alpine.js
• Reactive UI without SPA complexity
• End-to-end operational workflows for clinic staff
• Built for maintainability and future feature growth

Results

• Clinic operations consolidated into one coherent system
• Staff work in a fast, reactive interface instead of paper trails
• Seamless functionality across the clinic's daily workflows

Stack

Browser Extension · Scraping

E-commerce Data Extraction Extension

The problem

A client needed complete product catalogs — titles, prices, descriptions, specs, images, categories, URLs — from e-commerce sites, imported into their own platform. Copy-pasting was hopeless, and dynamic HTML broke naive scrapers.

The solution

I built a Chrome extension that extracts product data directly from the live DOM, normalizes it into structured JSON, opens product pages in parallel tabs for concurrent scraping, walks pagination automatically, and POSTs the collected dataset to the client's endpoint — supporting single and bulk product imports with titles, descriptions, images, and categories.

How it works

• DOM-based extraction of titles, prices, specs, images, and URLs
• Single and bulk product import to the target platform
• Parallel tab processing for concurrent page scraping
• Automatic pagination handling across result pages
• Throttled, asynchronous execution that stays stable at volume
• Defensive checks that filter incomplete or malformed records

Results

• Full product catalogs collected in hours instead of weeks
• Data arrives clean and structured — no post-processing needed
• Re-runnable at any time to keep the dataset current

Stack