Skip to content
AI automation specialist Manila, Philippines · remote worldwide

I build automations that give your business hundreds of hours back.

I connect the tools your team already uses, remove the repetitive data entry, and build AI systems with clear fallbacks, alerts and human approval where it matters.

100+ hrs
potential monthly time saved
5–7 days
to a working prototype
< 24h
typical reply time

01 · Focus areas

What I automate

Four places where repetitive work disappears fastest.

02 · Production workflows

Systems already running

Three automations in production, shown the way I would hand them over to you.

01 Daily Gmail executive briefing

Every morning at 7:00 AM, Asia/Manila

n8n canvas for the daily Gmail executive briefing workflow

The workflow pulls the last 24 hours of Gmail, strips out HTML boilerplate and signatures, and asks GPT-4o for a briefing: what needs attention, what changed, and what to do next. It lands in Telegram before the workday starts.

What it replaces

Founders lose one to two hours every morning triaging noisy email before they get to real work.

How it works

  1. 01 Schedule trigger fires at 7:00 AM
  2. 02 Gmail: fetch every message from the last 24 hours
  3. 03 Clean and normalise the raw email bodies
  4. 04 Branch: were any emails found?
  5. 05 AI agent writes the briefing with GPT-4o
  6. 06 Telegram: deliver the briefing to the founder

Safeguards

  • Email text is treated as untrusted data, never as instructions to the model
  • HTML stripped and long bodies truncated to keep token cost predictable
  • Temperature held at 0.2 so summaries stay factual and repeatable
  • A 3,500-character ceiling so Telegram never splits the message

02 Jibble to Google Sheets payroll preparation

Every Monday at 8:00 AM, Asia/Manila, or run by hand

n8n canvas for the Jibble to Google Sheets payroll workflow

Approved and locked pay periods come out of Jibble, ISO 8601 durations become payable hours, gross pay is calculated, and idempotent rows are written to Google Sheets. A manager approves by email before anything is final.

What it replaces

HR and finance teams spend hours cross-checking timesheets into spreadsheets, which invites duplicate pay runs and calculation errors.

How it works

  1. 01 Schedule trigger opens the prior week
  2. 02 Read pay rates and the employee list from Sheets
  3. 03 Fetch approved pay periods from the Jibble API
  4. 04 Parse durations and calculate gross pay
  5. 05 Upsert the rows into the payroll ledger
  6. 06 Email the manager and wait for a decision

Safeguards

  • Never initiates payment or statutory deductions, by design
  • Idempotent writes make a retry a safe no-op instead of a duplicate pay run
  • Strict duration parsing fails fast before anything is written
  • OAuth2 scoped access for every integration, with no hardcoded credentials

03 On-demand Reddit sentiment analysis

On demand, with parameters

n8n canvas for the Reddit sentiment analysis workflow

Point it at a keyword and a subreddit. It cleans the comment threads, filters deleted and NSFW posts, then scores the community's sentiment into positive, neutral, skeptical and critical clusters before reporting to Telegram.

What it replaces

Tracking customer feedback and competitor complaints across thousands of comments takes hours of manual browsing and ends in subjective conclusions.

How it works

  1. 01 Workflow inputs take the keyword and subreddit
  2. 02 Validate and sanitise every parameter
  3. 03 Search Reddit for matching threads
  4. 04 Fetch and clean the comment threads
  5. 05 Classify sentiment with the LLM
  6. 06 Telegram: send the market intelligence report

Safeguards

  • User comments are quarantined as untrusted evidence in structured blocks
  • Temperature 0.2, a 700-token cap and a 90-second timeout
  • Strict length and regex validation on all input parameters
  • A pinned sample mode for testing without spending tokens

03 · Services

What I can build for you

End-to-end systems designed around your actual tools and approval rules.

04 · Stack

Tools & infrastructure

A practical stack for reliable integrations, AI processing and monitored deployment.

Orchestration

  • n8n
  • FastAPI
  • Node.js / Express
  • LangChain
  • Docker
  • REST & webhooks

Models & perception

  • OpenAI GPT-4o
  • Claude 3.5 Sonnet
  • Gemini embeddings
  • Agora Voice AI
  • Whisper
  • Tesseract.js

Data & messaging

  • PostgreSQL
  • Supabase
  • Telegram Bot API
  • Meta Messenger API
  • Google Sheets
  • Gmail API
  • Reddit API

05 · References

What collaborators say

From the engineers and founders I've built with.

“Miguel is an outstanding engineer who consistently delivers high-quality systems. His work optimizing our database pipelines by 10x and implementing caching cut our latency by 90% while serving 500+ daily commuters seamlessly.”

Joshua Bumanlag

Senior iOS Engineer at ING Hubs · Creator of Rush PH

“Miguel's AI know-how and speed of execution have brought immense power to every project we've tackled together. He understands both the engineering architecture and the practical automation that businesses actually need.”

EJ Gungon

Senior Developer at Springboard Philippines

06 · Free process audit

Start with one slow workflow

Bring me a single repetitive process. I'll show you the smallest useful system that fixes it.

1

Map the bottleneck

We find the repetitive task with the clearest return, and agree where a human must stay in control.

2

Prototype in 5–7 days

I connect a safe sandbox to your actual tools and validate it against real examples.

3

Roll out and monitor

The production system ships with error alerts, fallback logic, rate-limit protection and documentation.

Tell me what's eating your week

Email me the process and I'll reply with how I'd automate it, what it takes and what it would cost. I usually answer within a day.

mgkrdklw26@gmail.com