Atlas Deha Coşan · Istanbul, Türkiye

Building AI systems, data products, and full-stack software.

Computer Science & Engineering student at Sabancı University, building RAG workflows, recommender systems, optimization models, analytics pipelines, full-stack platforms, and mobile applications.

AI EngineerSoftware Engineering InternFounder & OperatorCS @ Sabancı UniversityIncoming Intern Engineer @ FlyRank.aiIncoming Intern Engineer @ TEB
View ProjectsContact MeGitHubLinkedIn

About

Engineer, founder, builder.

Atlas Deha Coşan is a Computer Science & Engineering student at Sabancı University with experience across AI engineering, data analysis, backend systems, full-stack products, mobile development, and fintech infrastructure. He has built production-ready RAG workflows, recommendation systems, optimization models, analytics pipelines, and commercial digital systems through both technical projects and real business operations.

Code on a laptop screen
Analytics dashboard with charts
Modern office workspace

What I build

AI workflows and RAG systems
Data analysis and analytics pipelines
Recommender systems
Optimization models
Full-stack web applications
Mobile applications
Database-backed systems
Automation and business tools

Experience

From AI labs to fintech floors.

Internships, contracts, and a registered company — engineering experience built across AI training, banking technology, capital markets, and real business operations.

  1. FlyRank.ai logo

    FlyRank.ai

    Intern Engineer

    IncomingIncoming · Accepted

    Internship

    Accepted as an Intern Engineer at FlyRank.ai, expanding hands-on experience in AI-driven engineering, product systems, and applied software development.

    AI EngineeringProduct SystemsApplied Software Development
  2. TEB logo

    TEB

    Intern Engineer — Mandatory Internship

    IncomingIncoming · Accepted

    Internship

    Accepted as an Intern Engineer at TEB for the mandatory internship, gaining exposure to engineering practices in banking, financial technology, and enterprise software systems.

    Banking TechnologyFintechEnterprise Software
  3. Alignerr logo

    Alignerr

    AI Trainer

    CurrentFebruary 2026 — Present

    RemoteContract

    Works on AI training and evaluation tasks, applying artificial intelligence, data analysis, quality control, and RAG-related workflows to improve AI system outputs.

    Artificial IntelligenceData AnalysisRetrieval-Augmented GenerationQuality Control
  4. A

    Atlas Cosan LLC

    Founder & Operator

    CurrentFebruary 2025 — Present

    United States · RemoteSelf-employed

    Founded and operates Atlas Cosan LLC, a registered limited liability company based in Albuquerque, New Mexico. Built and distributed digital products, implemented data-driven marketing systems, SEO/SEM strategies, and automated sales workflows. Generated $8,000 in revenue.

    EntrepreneurshipBusiness DevelopmentDigital MarketingLLCAutomation
  5. Outlier logo

    Outlier

    AI Trainer

    CompletedFebruary 2025 — August 2025

    RemoteFreelance

    Worked as a freelance AI trainer, contributing to AI evaluation and improvement workflows with a focus on data analysis, quality control, and model response quality.

    Artificial IntelligenceData AnalysisQuality Control
  6. Tacirler Yatırım Menkul Değerler A.Ş. logo

    Tacirler Yatırım Menkul Değerler A.Ş.

    Software Engineering Intern

    CompletedAugust 2024

    Istanbul, Türkiye · On-siteInternship

    Helped develop and implement an SQL database using C# and .NET for a web-based application, improving data management efficiency. Set up both virtual and physical servers to support application deployment.

    JIRASQL.NETC#Database DevelopmentAgile Project Management
  7. TÜBİTAK BİLGEM logo

    TÜBİTAK BİLGEM

    Software Engineering Intern

    CompletedFebruary 2024

    On-siteInternship

    Designed modular and reusable code architectures using advanced C++ principles, strengthening object-oriented programming and software architecture fundamentals.

    C++Object-Oriented Programming

Featured Projects

Systems with measurable results.

Case studies across optimization, machine learning, and data science — each with a clear problem, a real build, and a quantified outcome.

Graduation Project · Optimization

EV Routing with Charging While Driving

Problem
Electric vehicle fleets lose time and range planning around static charging stops. Electrified Road Systems (ERS) allow charging while driving — but routing with dynamic charging and battery constraints is a hard combinatorial problem.
What he built
Developed a MILP formulation for the Electric Vehicle Routing Problem with dynamic ERS-based charging, incorporating battery state-of-charge constraints. Cross-compared routing algorithms using Python/Gurobi exact solving and an ALNS metaheuristic.
Result
A working exact-vs-heuristic benchmark: Gurobi delivers provably optimal routes on small instances, while ALNS scales to larger networks with near-optimal solutions.
  • MILP model with battery state-of-charge constraints across ERS segments
  • Exact solving with Python + Gurobi
  • ALNS metaheuristic for large-instance scalability
  • Cross-comparison framework for solution quality vs. runtime
PythonGurobiMILPALNSEVRPOptimizationAlgorithms

Code & report available on request

Machine Learning · Recommender Systems

Song2Vec Music Recommendation System

Problem
Music platforms need to recommend the next song from sparse listening sessions — collaborative filtering struggles with sequence context and cold-start items.
What he built
Built a music recommendation system using Word2Vec-style embeddings on the 30Music dataset. Implemented a Skip-gram model in Python with Gensim, tuned hyperparameters, used negative sampling and training epochs, improved HR@10 from 2.84% to 7.20%, and designed a full evaluation pipeline with cold-start analysis.
Result
Top-10 recommendation accuracy improved 2.5× over the baseline configuration after embedding and hyperparameter tuning.
  • Skip-gram embeddings over listening sessions (songs as tokens)
  • Negative sampling + epoch tuning for training stability
  • Systematic hyperparameter search across window size and dimensionality
  • Cold-start analysis for unseen tracks
PythonGensimWord2VecEmbeddingsRecommender SystemsML EvaluationHyperparameter Tuning

Code & report available on request

Top-10 hit rate2.84% → 7.20%

Data Science · Analytics

Data-Driven YouTube Algorithm Performance Analysis

Problem
Does creative effort actually move the needle on YouTube performance, or does the algorithm dominate? Atlas connected his own music channel — 137k views, 535 subscribers — with rigorous data science to find out.
What he built
Analyzed whether creative effort affects YouTube performance by combining YouTube Analytics / YouTube Data API data with time-tracking data from music production. Used pandas, regression analysis, hypothesis testing, visualizations, and p-value analysis to evaluate relationships between production effort and engagement.
Result
A statistically grounded read on how effort, creativity, and algorithmic performance interact across 137k views of real channel data.
  • Joined YouTube Data API metrics with personal time-tracking data
  • Regression analysis + hypothesis testing with p-value interpretation
  • Engagement visualizations across upload history
  • Real production data, not a toy dataset
PythonPandasYouTube APIData AnalysisRegression AnalysisHypothesis TestingVisualization
View on GitHub
Views analyzed137k

Full-Stack Web Application

Full-Stack E-Commerce Clothing Platform

Built a full-stack e-commerce clothing platform using React, Node.js, and MongoDB in an Agile team. Implemented Docker and docker-compose with RESTful API architecture.

ReactNode.jsMongoDBREST APIDockerDocker ComposeAgileScrum

Mobile Application

Cross-Platform Food Delivery App

Engineered a cross-platform food delivery mobile app for iOS and Android using Flutter and Dart. Created Figma UI/UX prototypes, followed Agile sprints, integrated real-time data handling, and designed with CI/CD compatibility in mind.

FlutterDartFigmaMobile DevelopmentUI/UXAgileCI/CD

Skills

Technical stack, backed by usage.

Hover a category to see where these skills were applied — every group maps to real experience or shipped projects.

AI / Machine Learning

Artificial IntelligenceLangChainHugging FaceOpenAI APIChatGPT APIRetrieval-Augmented GenerationRAG workflowsMulti-agent pipelinesRecommender systemsWord2VecHyperparameter tuningRegression analysisMachine learning evaluation

Used in

  • Alignerr — AI Trainer
  • Outlier — AI Trainer
  • Song2Vec recommender
  • DeepLearning.AI / Stanford certification

Data

Data AnalysisVeri AnaliziPandasHypothesis testingYouTube APIDatabase systemsSQLAnalytics pipelinesQuality Control

Used in

  • YouTube algorithm analysis — 137k views
  • Alignerr & Outlier QC workflows
  • Tacirler — SQL databases

Backend / Databases

Node.jsREST APIsC#.NETSQLMongoDBDatabase DevelopmentDatabasesServer setup

Used in

  • Tacirler — C#/.NET + SQL + server setup
  • E-commerce platform — Node.js + MongoDB

Frontend / Product

ReactJavaScriptFigmaResponsive UIComponent architectureFull-stack development

Used in

  • E-commerce platform — React
  • Food delivery app — Figma prototypes

Mobile

FlutterDartiOS & Android developmentReal-time data handling

Used in

  • Food delivery app — Flutter/Dart, iOS & Android

DevOps / Workflow

GitDockerDocker ComposeKubernetesDevOpsJIRAAgileScrumAgile Project ManagementAgile & Waterfall Methodologies

Used in

  • E-commerce platform — Docker + Compose
  • Tacirler — JIRA + Agile

Programming

PythonC++JavaScriptDartSQLObject-Oriented Programming

Used in

  • TÜBİTAK BİLGEM — advanced C++ & OOP
  • EV routing — Python/Gurobi
  • Song2Vec — Python/Gensim

Business

EntrepreneurshipBusiness DevelopmentDigital MarketingSEO/SEMLLC operationsAutomated sales pipelines

Used in

  • Atlas Cosan LLC — $8K revenue, automated sales workflows

Languages

Turkish — NativeEnglish — Full professional proficiency

Metrics

Numbers, not adjectives.

Experience spanning AI, fintech, data science, full-stack, mobile, and optimization — measured where it counts.

$0K

Revenue generated through Atlas Cosan LLC

0.0×

Recommendation accuracy uplift in Song2Vec

0k

YouTube views analyzed with regression & hypothesis testing

0

YouTube subscribers analyzed

0+

Major engineering projects

0

Incoming / accepted engineering internships

Education

Computer science, two continents.

Sabancı University logo

Sabancı University

Bachelor's Degree, Computer Science

October 2022 — January 2027

Istanbul, Türkiye

Data StructuresDatabase SystemsMachine LearningAgile & Waterfall MethodologiesData Analysis
City University of Hong Kong logo

City University of Hong Kong

Bachelor's Degree, Computer Science

September 2025 — December 2025

Hong Kong SAR

AlgorithmsOperating SystemsComputer NetworksData Analysis

Certifications

DeepLearning.AI · Stanford University logo

Supervised Machine Learning: Regression and Classification

DeepLearning.AI · Stanford University · Issued February 2025 · Credential ID 2K7XOBNQU7DS

Data AnalysisRegression AnalysisMachine Learning

Beyond the keyboard

Nursing Home Volunteer

Sabancı University Civic Involvement Projects · September 2022 — January 2023

Assisted elderly residents with daily activities and provided companionship.

Contact

Let's build something technical, useful, and sharp.

Open to AI engineering, software engineering, data science, full-stack, and fintech internship opportunities.

cosanatlas@gmail.comLinkedInGitHub