Micael Targino

Micael Targino

Based in Brazil, working remotely. Available for projects.

Lead Software Engineer.I ship systems that move numbers.

Backend, data and AI products running in production: ad serving, ETL pipelines, compliance automation.

  • Lead at Metrike
  • Ex-SnapMagic (YC S15)
  • 4+ years in production
Metrike Ad Server

Selected work

Five products I built or led, all of them shipped. Each one opens into the full story.

Metrike Ad Server

An in-house ad server for display, video and apps, built to replace a third-party platform.

Metrike Ad Server screen

Lead engineer

Laravel, FastAPI, Pub/Sub, BigQuery, Kotlin, Swift

Read the case study
−41% time to first ad on broadband
~5 ms server-side ad decisioning
200M+ ad events a day, scaling horizontally

Data pipelines and warehouses

ETL rebuilds and BigQuery migrations, measured before and after.

Data pipelines and warehouses screen

Led at Metrike and SnapMagic (YC S15)

Airflow, BigQuery, Python, Pandas, Airbyte

Read the case study
16 h 1 h full ETL run
4 h 20 s one critical task
~190 s ~2 s average report, ~50 queries

Media plan to live campaign

An AI agent reads agency media plans in any layout. An automation then builds the campaign on the ad server.

Media plan to live campaign screen

Design and implementation

Claude tool use, Laravel, Queues, REST APIs

Read the case study
~3 h ~10 min campaign setup
−83% time from media plan to live campaign

FaturaDoc

Checks that media suppliers are in good standing with tax and labor authorities before an agency pays them.

FaturaDoc screen

Took it over, rebuilt it, run it in production

FastAPI, Celery, PostgreSQL, React, Gemini

Read the case study
5 government certificates checked per supplier
−10k lines of dead code removed

MRF

A platform for Brazilian fixed-income investors: market data, yield math and portfolio tracking.

MRF screen

Freelance lead engineer, Oct 2024 to Jun 2025

FastAPI, Next.js, Airflow, Selenium, Redis

Read the case study
  • Market data scraped continuously through Airflow
  • IRR engine that handles business days, inflation and amortization
  • Server-rendered Next.js on GCP and Vercel

Inside Metrike

Smaller changes across the ad server, the brand-safety pipeline and the campaign platform. Each one measured.

  • 76%91%

    Brand-safety classifier accuracy

    Benchmarked four models on an adjudicated gold set, then migrated the AI layer. About 5× faster per URL.

  • ~50%74%

    Agreement on brand-suitability risk

    Designed the GARM risk dimension and tuned criteria and prompts against a two-judge consensus.

  • 30.6%

    Silent data loss, found and fixed

    Scraped pages were arriving empty with no error. The cause was missing Brotli decoding.

  • ~30 s~5 s

    Worst-case dashboard load

    Moved verification and location queries from MySQL to BigQuery.

  • 48 KB17 KB

    Ad tag bundle

    And the tag requests the ad in parallel with the consent check.

  • 2 SDKs

    Native Android and iOS ad SDKs

    Kotlin and Swift. MRAID 3.0, VAST video, viewability and conversion attribution, with public docs.

  • −75%

    Memory on the Excel report API

    Fixed the leaks in a Django, Celery and S3 service that builds large files.

Experience

  • 2026 – nowLead Software EngineerMetrike
  • 2025 – 2026Software Engineer IISnapMagic (YC S15)
  • 2024 – 2025Freelance lead engineerMRF
  • 2022 – 2025Software Engineer I and IIMetrike
  • 2023 – 2027B.Sc. Computer Science, in progressUFPB