index.html README.md career.log skills.yaml certs.json

Dustin Smith

Director Data Engineering, SRE & Infrastructure

const yearsExp = 10+

const countries = 5

const cloudSavings = "$3M"+

I’m Dustin—an engineer who’s spent the last decade building data platforms, ML systems, and SRE practices across five countries. I love the challenge of taking something broken or non-existent and turning it into a well-oiled machine. Currently Director Data Engineering, SRE & Infrastructure at Techcombank, where I built the SRE function from zero, lead a team of 20, and ship agentic tooling—an Ops Agent triaging pipeline failures in production and a CR agent that cut change reviews from hours to minutes—for one of Vietnam’s largest banks.

export const achievements = [

Built SRE function 0-to-1—leads a team of 20 across Reliability/SRE and platform infrastructure
Ops Agent in production—delivered first-ever reliability baselines (previously unmeasured): ~30 min MTTA and ~60 min MTTR; triages, groups, and explains job failures, with an auto-resolve engine in beta testing
Delivered CR agent CLI cutting change reviews from 90–120 min to ~10 min, sustaining ~50 CR reviews per week; developing agentic version with Confluence-to-GitLab automation
$1.03M AWS savings through lifecycle optimization, division-wide job right-sizing (25–50% resource reduction), and pipeline refactoring
Delivered observability across ~6,000 data pipelines with dashboards and alerting

Grew account 4x—built customer confidence and expanded spend from $200K to $800K annually
Landed $1M expansion by bringing Databricks to a second team at another customer
Hands-on code optimization cutting customer pipeline runtimes 37.5–60% through cluster right-sizing and Spark tuning

Built the MLOps team and established practices for model lifecycle management
Integrated ML Metadata (MLMD) with Vertex AI for tracking training and serving data
Drift detection pipeline to monitor model performance degradation in production

2.331 PB data optimization on-premise, saving 40M THB (~$1.2M) in storage costs
Led cloud migration evaluation and POC from on-prem to GCP + Databricks
Established data engineering standards that scaled across multiple teams

];

3181045

Your Parquet Column Indexes Are Being Ignored on EMR and Glue

Aug 10, 2026

893a282

Interlock: A STAMP-Based Safety Framework for Data Pipelines

Feb 24, 2026

e3d4a1e

PySpark Pipeline Framework: Configuration-Driven Pipelines for the Python Ecosystem

Feb 7, 2026

739c5d3

Hardening Gastown: Role-Based Access Control for Multi-Agent Workflows