Preslav – React, Node.js, Typescript

In his role as CTO, Preslav took charge of the critical aspects of software development. This involved making strategic technology decisions, designing system architectures, optimizing data flows, upholding coding standards, overseeing multi-tenancy solutions, shaping administrative interfaces, creating and sticking to a development roadmap, managing risks, and establishing core infrastructures. Through these efforts, he ensured the effective fusion of technical prowess with overall business objectives, driving the success of the software projects under his purview.

• defined tech stacks and project approaches;
• designed the system architecture and defined the data models;
• created of project roadmaps;
• assessed risks.

This project involved building a robust data integration platform designed to connect and stream data from various third-party providers, including SumUp, Stripe, Notion, custom APIs, and others. The platform was engineered to support real-time and batch data processing using technologies such as Bytewax, Faust streaming, Apache Flink, and Apache Beam.

• Developed and deployed a comprehensive integration dashboard for monitoring and managing data pipelines.
• Integrated multiple data sources, ensuring smooth ingestion and processing from diverse providers like SumUp, Stripe, Notion, and custom APIs.
• Designed and implemented deployment workflows using Kubernetes, leveraging Helm charts for scalable and maintainable infrastructure setup.

The platform ensures that only safe, high-quality, and relevant data is fed into language models. The labeling technology allows companies to easily:

1. Auto-generate metadata describing the contents of their data
2. Find & remove sensitive information
3. Control what data can be accessed by which user

• implemented RAG Strategies (Recursive Retriever, RouterQueryEngine, SubQuestionGenerations, AutoVectorIndexRetriever);
• added RAG Vector Index providers integration using LlamaIndex (OpenSearch, Azure AI, Qdrant, Postgres);
• generated Metadata for RAG pipeline;
• provided RAG evaluation;
• conducted stress testing of the RAG System;
• implemented RAG API.

HTML banner generation using Open AI and Stability AI.

• created front-end using Next.js Page Router;
• built back-end using Node.js and Prisma ORM;
• deployed the solution using Vercel, Digital Ocean, and AWS;
• created Open AI using GPT 3.5 and Stability AI integration.

Chatbot for discovering information in manufacturing documents and answering questions with Open AI.

• created front-end using React with Fluent UI;
• built back-end using Python, Flask, and Azure SQL;
• deployed the solution in Azure App Services;
• integrated Open AI with GPT 3.5 and GPT 4 combined with Azure Search AI.

Smart, customer-centric software solutions that empower Shippers and Carriers, which results into an unified auto transport industry software with fewer hassles, zero paperwork and faster shipments.

• planned architecture and features like Create/Edit Screens, Dashboard, Reporting, Listing Screen, Users Screen, Profile Screen);
• led the development effort (Distributed and planned work, Planned Features, and Tech Debt);
• led a team of developers and broke down work-related tasks;

• communicated with the Product Owner (Scrum Process, Requirement Refinement).

Global interactive database to improve understanding of cesarean section rates and outcomes using the Robson classification system. It is meant to be used by doctors. The main Features are Logging data for cesarean section rates.

The system was later expanded by incorporating and AI model to read through complex data and descriptions in order to figure out relevant results.

The platform executes complex information queries using large datasets, human-readable documents and specific healthcare databases such as FDA, MAUDE, NHS, Adverse events database

• designed end-to-end solutions;
• developed features like Home Pages, Listing Pages, Charts, Details Pages, Create/Edit Pages, Profile Pages;
• developed back-end features like Search API and CRUD API;
• built deployments (Using Docker);
• incorporated ChatGPT 3 and 3.5;
• built and tested prompts for several models, experimenting with temperature and proper context settings.

Propaganda Detection in multilingual memes.

• performed data crawling;
• gathered, annotated, and analyzed meme data;
• classified the data by using and creating new multimodal models;
• conducted data analysis;
• did data annotation;
• modeled the architecture;
• created models.