Machine Learning + Data + Systems Engineer

AV monogram

Aris Vlasakakis

Reliable ML systems for real-world scale.

You get product-ready inference, dependable data pipelines, and pragmatic engineering practices that raise velocity without trading reliability. I favor functional programming because testability is a core requirement for how I design systems.

About

You’re working with a hands-on engineer focused on measurable outcomes: reliability, cost, and velocity. I build ML systems that support $1M+ in daily revenue and keep teams shipping safely.

Principles

Testability Reliability Evolvability

Impact Snapshot

-90%

Inference infra cost after moving to Triton.

-75%

Sev 0 incidents with testing hardening.

+40%

Team velocity after modernizing practices.

$1M+

Daily revenue supported by ML systems.

Core Expertise

ML Engineering

  • Inference systems with Triton and custom serving layers
  • Model testing infrastructure and reliability guardrails
  • Open-source LLM integration and deployment
  • Batch recommender systems at massive scale

Data Engineering

  • Large-scale learning pipelines in SparkML
  • Batch recommender pipelines in Dataflow
  • Streaming dataflows and Kafka-driven systems
  • Airflow migration and orchestration modernization

Software Engineering

  • Composable architectures with Scala + Python
  • Performance, latency, and cost optimization
  • Team enablement and sustainable delivery

Selected Work

Large-Scale Learning on SparkML

Built distributed training pipelines to improve throughput and model iteration speed.

Stack: SparkML · Scala · Python

Batch Recommender Systems

Built user-product recommenders for tens of millions of users and billions of items.

Stack: Dataflow · TensorFlow · PMML

Inference Migration to NVIDIA Triton

Replatformed serving to reduce latency and cut infra costs by 90%.

Stack: Triton · Docker · Monitoring

Inference Testing Infrastructure

Designed reliability tests that dropped Sev 0 incidents by 75%.

Stack: CI/CD · Python · Observability

Modernized Model Orchestration

Upgraded orchestration to accelerate releases and enable LLM workflows.

Stack: Airflow · Workflow Orchestration · LLMs

Systems & Stack

Scala Python SparkML Dataflow Kafka Triton TensorFlow PMML LLMs Docker Airflow CI/CD Observability

Working Style

You get a pragmatic builder who obsesses over reliability, clear contracts, and measurable outcomes. Systems ship with observability, tests, and an upgrade path baked in.

Aris Vlasakakis

Machine Learning + Data + Systems Engineer

Let’s Build Something Durable

Reach out for ML systems, data platforms, or reliability-focused engineering leadership.

Connect on LinkedIn