Stanislav Yaranov

An independent SaaS platform for tutors. The goal is to consolidate lessons, students, learning materials, interactive exercises, and routine administrative work into a single system rather than relying on several disconnected tools.

I am responsible for both product development and technical implementation: defining the MVP, working through user workflows, designing the architecture, and building the application end-to-end. The platform uses Next.js and React for the front end, NestJS with Express for the API, PostgreSQL for data storage, and a separate Python/FastAPI service for AI-related functionality.

The product is being designed around interactive lessons, reusable workbooks and exercises, student management, content libraries, and AI-assisted workflows such as lesson preparation, transcription, summarization, and content generation.

A personal website that includes an AI chat for answering questions about my experience, projects, and technical background. The system is split into a Next.js web application and a Python and FastAPI AI service using OpenAI, LangGraph, PostgreSQL, and pgvector. It handles message validation, intent classification, retrieval, fallback cases, and hiring-intent detection. It is deployed on Google Cloud.

Designed and implemented a production-shaped LLM/RAG application for conversational interaction with YouTube video content.

The system accepts YouTube URLs, extracts or transcribes video content, chunks transcripts, stores semantic embeddings in PostgreSQL using pgvector, and provides retrieval-scoped conversational chat for each processed video context.

Architected the application with clear service, repository, gateway, and API boundaries using FastAPI and LangChain. Implemented asynchronous ingestion workflows with Celery and Redis, including retry-aware background processing, stale job recovery, idempotent ingestion keyed by YouTube video ID, and cleanup of vector records on persistence failures.

Developed a React/Vite front end supporting video library management, streaming chat responses, and context switching between processed videos. Added focused unit tests for ingestion, retry lifecycle, retrieval scoping, and high-risk workflow behavior.

Built as a practical exploration of modern AI-enabled application architecture with emphasis on reliability, maintainability, and production-oriented engineering patterns.

Contributed to the evolution of a large-scale AI-powered branding platform focused on visual content and brand asset generation.
Designed and implemented the core architecture of a complex Canva-style editor with custom state management, geometry processing, rendering logic, and React integration layers optimized for heavy, real-time interaction and performance-sensitive workflows.
Integrated AI-assisted design workflows using tools such as OpenAI and Midjourney to improve asset generation and creative productivity. Also contributed across back-end and infrastructure layers, including optimization of serverless and microservice workflows to improve execution efficiency and reduce operational costs.

Architected and led development of a regional, mission-critical emergency response platform operating across the Republic of Tatarstan. The system processed realtime emergency calls, coordinated incidents between multiple emergency departments, and integrated telephony (Avaya), GIS infrastructure, vehicle tracking, and reporting subsystems.
Designed as a distributed high-availability platform with a strong focus on operational reliability, consistency, and real-time coordination under critical conditions. The system was used for day-to-day emergency response operations at a regional scale and required continuous availability across multiple agencies and services.

Designed and developed a regional real-time telematics platform for monitoring school buses, enforcing route compliance, and detecting speed violations across the Republic of Tatarstan.
The system processed live vehicle telemetry streams, integrated GIS mapping and dispatcher communication, and generated automated alerts and analytical reports for operational monitoring. Built around distributed back-end services and real-time event processing, the platform enabled large-scale day-to-day transportation oversight and improved operational visibility and safety.

Designed and led development of a centralized operational platform for the International Universiade 2013 event in Kazan, integrating arrivals, accommodations, incident management, and multiple operational subsystems into a unified real-time coordination interface.
The system provided operational visibility across venues and services, supporting schedule control, incident routing, and cross-department coordination during a large international event. Built to operate reliably under event-scale load with continuous, real-time data flow and coordination requirements.

Led architecture and full lifecycle development of a regional geospatial platform used by municipal and environmental departments across Tatarstan.
The system supported multi-layer spatial data modeling, geospatial editing, media integration, and cartographic reporting for operational monitoring and environmental analysis. Built as a scalable web-based GIS platform, it integrated large geospatial datasets and supported day-to-day analytical and operational workflows at the regional scale.

A custom embedded control system for smart HVAC airflow regulation based on STM32 microcontrollers. Designed and implemented firmware in C++ using an event-driven architecture for deterministic, real-time control of dampers and airflow balancing. Integrated hardware interfaces (I2C, RS-485/Modbus, sensors, actuators) and focused on reliability, performance, and system-level robustness under hardware constraints. Designed as part of a modular architecture integrating embedded control units with a higher-level supervisory system.

The Toptal React Accelerator is an exclusive learning program that educates JavaScript developers in React, NextJS, Redux, and GraphQL. This is a two-month program that consists of rigorous learning exercises and hands-on activities along with a stringent evaluation process, comprehensive learning assessments, and 30-40 hours of building an Invoice Management System. Within this invoicing application, users are able to track invoices, clients, and company information.