Hi, I'm Vytautas.

Built a layered dbt pipeline for Solvency reporting: source normalization, enrichment with submission/remittance metadata, latest-valid state resolution, and final analytics-ready indicator tables for regulatory analysis.

• Implements history-aware logic for latest and latest-valid submissions.
• Combines entity-level and group-level indicator calculations.
• Adds data tests and documented seed-driven indicator definitions.

The project solves the need to turn file-based source data into reliable, traceable analytical tables without losing historical context. I designed an Airflow-orchestrated ELT flow that backs up raw files, loads them into staging schemas, and uses dbt to build SCD2-style dimensional models with tests and documentation.

• Owned the backup → load → transform → test → document workflow across object storage, the analytical database, and dbt models.
• Preserved raw source files in dated backup areas before loading normalized staging tables.
• Added data quality checks and documentation so downstream users could understand lineage and table behavior.

The project solves fragmented regulatory reporting data by giving analysts one governed fact layer for consistent downstream use. I refactored the dbt SQL into readable, well-structured CTEs that consolidate raw records with metadata, entity attributes, and submission status.

• Clarified transformation logic so maintenance and review work could happen safely across the fact model.
• Kept enrichment rules explicit for metadata, entity attributes, and submission-state handling.
• Improved documentation value by making the model structure and business logic easier to inspect.

The CLI solves repeatable data maintenance for CSV backups, staging loads, and retention cleanup across S3 and Trino/Iceberg. I built a Python command-line tool with configurable backup, load, and cleanup modes driven by environment variables and CLI arguments.

• Implemented date-based S3 backup copies between accounts before data is loaded downstream.
• Normalized incoming CSV columns and wrote them into Trino *_staging tables.
• Added retention cleanup logic so old backup folders can be removed consistently.

Simple website deployment automation with GitHub Actions and Terraform.

A fully serverless contact form for my portfolio website. The frontend sends a POST request with user input (email and message). API Gateway acts as a secure HTTP interface, routing requests to the backend. AWS Lambda processes incoming requests, validates the payload, and triggers email delivery. Messages are sent via AWS SES — no servers, no long-running services.

The project solves property administration questions that are difficult to answer because building records, BIM data, and technical documentation are split across different sources. I created a RAG assistant prototype that ingests building passport records, normalized BIM tables, and PDFs, then uses n8n, OpenAI embeddings/chat, and ChromaDB retrieval to answer from retrieved context.

• Orchestrated ingestion and runtime retrieval across passport, BIM, and documentation sources.
• Grounded responses by identifying the building and retrieving context from the most relevant source before answering.
• Documented representative property-management questions and the runtime architecture for future extension.

RAG Runtime Architecture

Automated weather data ingestion to Postgres with subsequent analysis.

Pipelines for ML experiments and price prediction.

Prototype with data preparation and a web UI.

Containerised example with Airflow orchestration.