Custom Software Development

Creating solutions for healthcare and education: from DLL plugins to web platforms with fine‑tunable neural networks — turnkey.

Discuss a project

Services

ML for Healthcare

ML‑powered medical software: automatic ECG anomaly detection, organ segmentation on ultrasound/MRI, abnormality detection on scans. Convolutional neural networks (CNN), transformers, PyTorch/TensorFlow. DICOM integration, real‑time processing. Boosting diagnostic accuracy.

ECG Ultrasound MRI DICOM Tensorflow scikit-learn CNN

Web Services & APIs

Modern web applications in Python: high-performance backends with Django, FastAPI, Flask. REST, GraphQL, AsyncIO, database and cloud integration. Scalable, secure, suitable for any task from MVP to high-load systems.

Python Django Flask PostgreSQL MySQL Mongo Redis Docker

Desktop Applications

Professional desktop solutions in Delphi: native code, lightning-fast database access (FireDAC, ADO), perfect for ERP, accounting, CRM, and complex business logic. Supports Windows XP to 11.

Delphi VCL Windows MS SQL MS ACCESS

Automation & Chatbots

Smart Telegram bots with advanced prompt engineering: ChatGPT integration, business assistants, support automation, polls, notifications. Dialog scenario design, inline keyboards, payments, external APIs. Enhance customer service with AI.

Telegram API Prompt Engineering ChatGPT aiogram

Portfolio / Medical Projects

Web ECG explorer for bracelet

Web‑based explorer for short ECG recordings from a wearable bracelet. Uploads noisy signals, applies adaptive filtering (wavelets + IIR) and a CNN‑based neural network to automatically recognize cardiac events (normal, extrasystole, artifact). Implements user‑annotation fine‑tuning: the doctor corrects labels, the model updates online without full retraining. Interactive zoomable chart, report export, confidence scores.

React Python DRF Flask TensorFlow Scikit-learn Signal Processing Active Learning

DLL plugin for BLE with ECG/BP recorders

Universal DLL‑based communication module for BLE/USB interaction with a set of wearable ECG and blood pressure (BP) recorders. Implements measurement parameter programming, remote start, background reading with data buffering. Built‑in packet integrity checks and automatic reconnection on signal loss. Delivered as a plugin for a desktop application (Delphi / C++), easily integrated via exported functions.

Delphi C++ DLL BLE (Bluetooth Low Energy) Windows Medical Protocols

ECG Detection Quality Validation Service

REST API service for validating ECG signal detectors. Upload your recordings and the recognition module to test; the service compares its performance against ground‑truth annotated library records and provides a detailed report on discrepancies and overall quality metrics. Perfect for healthcare startups and research labs.

REST API ECG Signal Processing Validation Medical Software

Neural network system for anatomical structure segmentation

Neural network system for anatomical structure segmentation on MRI scans (bones, organs, lesions). A web resource for distributed DICOM data collection and labeling: radiologists and experts upload scans via browser, annotate contours using interactive tools, and generate ground truth. Role‑based system (admin, expert, labeler), quality control via consensus labeling, and export of annotations for training CNNs (U‑Net, Mask R‑CNN). The model fine‑tunes on new data to improve recognition accuracy.

Delphi MRI DICOM Medical Segmentation TensorFlow CNN Distributed Labeling

Portfolio / Educational Projects

Game‑based learning: Python, Java, C++

Engaging programming lessons through game development. Learn Python, Java or C++ by building your own games: from arcades to RPGs. Perfect for beginners and kids.

Python Java C++ GameDev

Test quality analysis based on IRT

Web‑based psychometric analysis tool based on Item Response Theory (IRT). User uploads a test item file and a student response matrix in a convenient format (CSV/Excel). The system automatically calculates item parameters (discrimination, difficulty, guessing) and latent abilities. Visualizes information functions, item characteristic curves, and test function. Generates a detailed PDF report with recommendations for test improvement.

Python IRT (Item Response Theory) Django Mongo PDF generation SciPy Rasch model Birnbaum model

Algorithmic Thinking Test for Applicants

A test for students enrolling in programming courses, assessing basic algorithmic skills before learning any programming language. Tasks include logic puzzles, pattern recognition, sequences, and simple algorithms. The result helps determine a student's readiness and recommend a personalized learning path. No syntax — just pure thinking.

Python Algorithmic Thinking Student Assessment

Philosophical Art Association Website

A collection of articles on philosophy and art with a unique publishing system. Authors write articles in RST (reStructuredText) markup, which automatically builds web pages and generates print‑ready PDFs. One source — two output formats, preserving semantic structure and style. Perfect for academic and creative communities.

Python RST (reStructuredText) Philosophy

About Me

I am a software developer focused on medical and analytical systems. I have experience building end-to-end solutions — from low-level plag-in module for wearable ECG and BP recorders to web platforms for psychometric analysis and neural network segmentation of MRI. I cover the full lifecycle: data collection and labeling (DICOM, distributed annotation), building ML models (CNNs, active learning, IRT), and integrating them into final products. I work with noisy signals, extract clinically relevant information, design modular plugins for desktop applications, and create interactive web interfaces for domain experts.

Tech Stack

Python Delphi C++ Javascript Java TensorFlow Scikit-learn CNN Postgres SQL MySQL MS SQL Mongo Redis DICOM BLE IRT Signal Processing Docker

Contacts

Email: contactszobin.org

Telegram: @szobin

WhatsApp: +52 312 199 72 05

Reach out any convenient way — let's discuss your project.