Amazon Web Services
AWS Certified Solutions Architect — Associate
Validates the ability to design distributed systems on AWS: this site's DNS, CDN, storage, compute and email layers are an applied example of that, not a study exercise.
PORTFOLIO — SHEET 01
Cloud & AI Engineer
Infrastructure should be code, not archaeology. I'm a Cloud & AI Engineer with a master's in Artificial Intelligence and AWS certified Solutions Architect — Associate. I build the AWS infrastructure that AI systems run on. This page is the proof: every resource provisioned with Terraform, integrated with GitHub Actions CI/CD, no console clicks.
Hover or tab through the diagram — each node explains the role it plays in serving this exact page.
Amazon Web Services
Validates the ability to design distributed systems on AWS: this site's DNS, CDN, storage, compute and email layers are an applied example of that, not a study exercise.
Amazon Web Services
Foundation-level certification covering AWS global infrastructure, core services, security, and cloud economics.
University of Michigan
Validated foundational proficiency in Python, including data structures and core programming logic.
LIVE
Statically hosted on S3, distributed through CloudFront with a free ACM certificate, routed by Route 53, with a serverless contact form on API Gateway, Lambda and SES. Infrastructure is fully defined in Terraform — nothing was clicked in a console.
MACHINE LEARNING
An end-to-end forecasting pipeline predicting wind and solar energy surplus 24 hours ahead. Built with temporal LSTMs and XGBoost, and integrated with a financial cost-minimization framework to safely offer free local electricity tariffs while mitigating spot-market deficit risks.
DEEP LEARNING
Master's thesis evaluating Vision Transformers (ViTs) against CNNs (VGG16, EfficientNetV2-B1) on the ISIC 2017 dataset. Implemented 5-fold cross-validation, custom data augmentations, and TFRecord pipelines. The ViT model achieved a peak ROC-AUC of 0.844.
MACHINE LEARNING
Kaggle competition pipeline predicting daily sales for 1,115 Rossmann pharmacies across Germany. A blended XGBoost + LightGBM ensemble with advanced feature engineering (target encoding, log-transformation) and a closed-store heuristic achieved a public RMSPE of 0.116 — top-tier on the leaderboard.