I automate data, decode problems, and build AI-driven solutions.

Dejan Stajic – Data Engineer & Problem-Solving Aficionado

View My Work
Dejan Stajic - Data Engineer

About Me

I'm Dejan Stajic, a data engineer based in Orlando, FL. I believe that well-structured data pipelines and clear processes eliminate chaos, one task at a time. My approach is simple: break problems into first principles, automate the repetitive, and deliver transparent solutions that scale.

Experience Highlights

  • 5+ years building and maintaining ETL/ELT pipelines with PostgreSQL, Airflow, and Python
  • Creator of DataPilot: an AI-driven DataOps platform for instant debugging, versioned sandbox environments, and automated ticketing workflows
  • Founder of Daily Debug Challenge: a website gamifying real-world debugging tasks to train developers
  • Built an RL-powered Yahtzee AI using Monte Carlo and PPO to demonstrate curriculum learning
  • Active contributor to open-source data engineering tools and community forums

Core Values

Clarity over Complexity

Every solution must be transparent—no hidden steps, no black boxes.

Automation with Purpose

If a task repeats more than twice, I automate it.

Truth-First Approach

I validate assumptions with data; if evidence is lacking, I generate and test hypotheses.

Spiritual Discipline

Rooted in Essene-inspired purity and moral integrity—keeping me focused, disciplined, and ethical.

Services & Expertise

What I do to solve your toughest data challenges

Data Pipeline Architecture

Design, implement, and optimize data workflows using Apache Airflow, dbt, and SQL.

  • • Zero-downtime schema migrations
  • • Fault-tolerant scheduling
  • • Performance optimization

Database Development

PostgreSQL cluster configuration and maintenance with real-time analytics.

  • • Cluster configuration
  • • Performance tuning
  • • Materialized views

AI/ML Experimentation

Reinforcement learning research with Monte Carlo, PPO, and curriculum learning.

  • • MLflow experiment tracking
  • • Distributed RL training
  • • Ray cluster deployment

Web Application Development

End-to-end development with React front-end and FastAPI back-end.

  • • Data-centric dashboards
  • • Docker + Kubernetes
  • • Scalable deployments

DevOps & CI/CD

GitOps workflows, automated testing, and continuous delivery.

  • • Infrastructure as Code
  • • Terraform deployments
  • • Reproducible environments

Problem Solving

First-principles approach to complex data challenges and system optimization.

  • • Root cause analysis
  • • Process automation
  • • Scalable solutions

Key Projects

Innovative solutions that demonstrate my expertise

DataPilot

In Development

AI-powered DataOps platform for instant debugging and automated workflows

Problem

Data analysts spend hours debugging failed ETL jobs, tweaking SQL, and waiting for downstream approvals.

Solution

  • • File uploads with auto-upsert vs. truncate into sandbox schema
  • • AI-generated SQL fixes with trend analysis
  • • Auto-push to production upon ticket approval
  • • Daily reminders and real-time backups
Impact: Reduced average ticket resolution time by ~60%. Alpha testers report a 40% increase in productivity.

Daily Debug Challenge

Launched April 2025

Gamified debugging platform for developers

Gamify debugging for junior and senior developers by presenting real-world ETL failures, SQL bugs, and Python errors.

Features

  • • Leaderboards for fastest correct fixes
  • • Automated hint system powered by LLMs
  • • Community-submitted puzzles with peer review

Tech Stack

FastAPIPostgreSQLReactTailwind CSSGitHub ActionsDocker
Results: Over 1,000 registered users in the first month; average session time of 18 minutes.

Yahtzee RL Agent

Completed

Near-optimal Yahtzee player using inverse reinforcement learning

Data Collection

Logged 500 games; performed IRL to infer reward function

Deterministic Phase

Trained supervised LLM policy on reduced-rule mini-Yahtzee

Stochastic Phase

Used Monte Carlo rollouts + PPO to fine-tune policy

Tech Stack

Ray ClusterMLflowPPOMonte Carlo
Results: Agent consistently scores in the 98th percentile against baseline heuristics.

AI-Driven Dog Emotion Detector

Prototype

Detect canine emotions from vocalizations and body posture

Inspired by Smokey the Keeshond & Gia the Pit Bull, this prototype detects dog emotions from audio and visual cues.

Method

  • • Collected 200+ hours of dog barks and videos
  • • CNN model for body pose estimation
  • • RNN for audio features
  • • Lightweight Streamlit web demo

Emotions Detected

HappyAnxiousAlertHungry
Results: Early prototype with ~75% accuracy across four emotion categories.

Skills & Tools

Languages

PythonSQLBashJavaScript

Data Engineering

Apache AirflowdbtPostgreSQLRedisAWSDockerKubernetes

Machine Learning

Ray RLlibTensorFlowPyTorchMLflowscikit-learnPandas

Web & DevOps

FastAPIReactTailwind CSSGitHub ActionsTerraform

Monitoring

PrometheusGrafanaELK Stack

Education

B.S. Computer Science, UCF (2018)

Google Cloud Professional Data Engineer (2021)

AWS Solutions Architect – Associate (2022)

Testimonials

"Dejan turned our manual dispatch logs into an automated system overnight. Tickets go from request to resolution in half the time."
MR
Michael Rodgers
CTO at Osceola Sod
"Daily Debug Challenge forced our team to level up. The puzzles mimic exactly the kinds of SQL edge cases I see daily."
SP
Sara Patel
Lead Data Analyst at FinTech Inc.

Let's Connect

Ready to solve your toughest data problems together?

dejan@datapilot.us
Orlando, FL (open to remote)
LinkedIn GitHub
"Let's solve your toughest data problems together. No jargon, no fluff—just results."
Get In Touch