Gustaw Max Szperliński

Data Engineering • AI & Automation • Digital Twins • Industrial Mechatronics

I’m passionate about technology and mechanics. My experience includes working on digital twins for vaccine manufacturing processes, specifically to support the optimization of parameters in protein bioreactors. Bridging real-world industrial processes with emerging technologies is the field I want to pursue further. This is why I transitioned into a new industry and now work at a company specializing in the integration of biomass production lines. In November, I will also begin a postgraduate program in Industrial Mechatronics at Wrocław University of Science and Technology, allowing me to combine my existing IT experience with practical knowledge of machinery and physical processes.

Featured Projects

  • A VIN-centric platform for monitoring the sports car market and tracking vehicle valuations over time. The relational data model is built around a master table of unique VINs linked through a one-to-many relationship to an observations table, enabling the tracking of individual vehicles throughout their lifecycle and the analysis of price fluctuations over time. The pipeline decodes data directly from Otomoto's hydrated data streams by parsing __NEXT_DATA__ JSON objects and GraphQL requests, while unstructured vehicle attributes are normalized using an LLM.

    Python VIN-Centric Schema (1:N Relations) Otomoto __NEXT_DATA__ Decoding GraphQL / Hydration Parsing LLM Attribute Extraction Time-Series Price Tracking
  • An automated monitoring system for Yamaha motorcycle listings on OLX, built on a serverless architecture. The system periodically queries OLX's internal REST endpoints to identify motorcycles matching specific model and engine criteria, retrieves detailed listing metadata, and submits listing images to the Gemini 2.5 Flash Vision multimodal model. The model analyzes the images for Yamaha's distinctive factory Iconic Blue finish across body panels, wheels, and other exterior components, automatically filtering out listings that do not meet the specified visual criteria.

    Cloudflare Workers Gemini 2.5 Flash Vision OLX Internal REST API Serverless Cron Triggers KV Storage
  • Asystent Menago (Menago AI CRM) Commercial System

    A dedicated AI-native operational assistant and CRM system built for an artist manager. It uses retrieval-augmented generation (RAG) and vector embeddings of historical data—including invoices, project budgets, contracts, and plans—to provide the assistant with rich operational context for analysis and question answering. The platform automates document workflows and invoicing, using Gemini 2.5 Flash Vision for invoice OCR and the Polish Bielik model for chat and command handling. A key architectural safeguard is its Human-in-the-Loop approach: all proposed changes to records, budgets, or documents are routed to a dedicated approval queue, where the manager personally reviews and authorizes each action before execution.

    RAG & Vector Embeddings Human-in-the-Loop Architecture Bielik LLM (Chat & NLP) Gemini 2.5 Flash (Vision OCR) Gmail & Google Workspace API TypeScript
  • An ETL pipeline for automatically collecting, parsing, and cleaning data on Polish wood-industry businesses from the public CEIDG and KRS registers. The system standardizes business entity records, geocodes addresses, and provides an interactive analytical dashboard with geospatial visualizations.

    Python CEIDG & KRS APIs Automated ETL Pipeline Geocoding & Spatial Data
  • A digital service and spare-parts ordering portal for GROSS industrial machinery. It uses native Web APIs to scan QR codes directly through device cameras, features an interactive catalog with a hierarchical machine component tree, and includes a dedicated administration panel for managing the spare-parts database.

    TypeScript Web QR Scanner API Hierarchical Parts Tree