Denys Dumynskyi - Full-stack developer delivering end to end software products | ContraWork by Denys Dumynskyi
Denys Dumynskyi
pro

Denys Dumynskyi

Full-stack developer delivering end to end software products

New to Contra

Denys is ready for their next project!

Cover image for Role-Based AI FAQ System: Right
Role-Based AI FAQ System: Right Answer for Every User, Instantly Preview description: We built a modular, AI-assisted FAQ platform where the LLM acts as an intelligent query router to a centralized knowledge base, reducing support workload by 40%–70% and providing public, investors, partners, and employees with accurate information. Problem The client is a manufacturer of advanced multifunctional materials. He wanted to ensure all interested users had access to relevant information quickly, reliably, and without placing an extra burden on the company’s experts. A conversational interface was required to enable users to ask questions in natural language and receive responses while keeping AI behavior controlled and transparent.  Solution We designed a standalone OpenAI-powered FAQ module with a controlled-response engine that generates answers based on predefined information, taking roles and permissions into account. As it is separated from the client’s Vercel-based website, the integration takes place via an iframe, while account management and access level assignment are handled within Vercel CMS.  When a user asks a question, the system filters the query based on the identified role (public, investor, partner, or employee) to eliminate data leakage. Then it applies the vector search to deliver an answer that is stored in the admin panel. The system also ensures that each response is accompanied by documents and links to Vercel pages based on available tags.  Results The FAQ system reduces the average time needed to find information from 10–15 minutes to under 30 seconds through conversational search and centralized knowledge management. The support workload decreased by 40%–70%. Onboarding time for employees, partners, and investors decreases by 50–80% while the system grants instant access to structured answers.  Tech stack: - OpenAI API for query understanding - VectorDB (pgvector) database for semantic question matching - PostgreSQL as a structured database for storing FAQs, tags, roles, and permissions - Python/FastAPI for backend and admin panel development - iframe for secure communication - Vercel CMS Vercel-hosted frontend integration
0
6
Cover image for Denbeaux: Legal Workflow Modernization Cut
Denbeaux: Legal Workflow Modernization Cut 69+ Hours of Manual Work to 5 Minutes Preview description: We modernized a legacy Excel-based legal workflow into a scalable web platform that detects fraud case matches, generates personalized letters, and sends them automatically, cutting 69+ hours of weekly manual work to 5 minutes. Problem A US-based legal consultancy used a manual Excel-based workflow to process fraud cases and contact potential claimants. The team spent more than 69 hours every week collecting raw case data, checking possible matches, preparing letters, correcting addresses, and mailing documents. The process worked for a limited scope, but it was slow, repetitive, and difficult to scale from automobile fraud cases to real estate fraud cases. Solution IT Craft transformed the client’s outdated Excel-based workflow into a scalable web application with cleaner architecture, automated case processing, and integrations with legal data sources and postal services. The system retrieves raw case data from third-party sources, enriches and validates it, detects relevant fraud case matches, and generates personalized PDF letters automatically. We also integrated PostalMethods API, added ZIP code and address correction, and built a real-time dashboard for tracking case status and letter delivery. The new architecture supports both automobile and real estate fraud cases, allowing the client to expand to new case types without rebuilding the workflow from scratch. Results Manual work dropped from 69+ hours per week to 5 minutes. The platform delivered 800x time savings, real-time fraud case tracking, automated PDF letter generation, postal dispatch, and expansion to new case types without additional development. Tech Stack: .NET 8 Angular 16 AWS IronPDF PostalMethods API
0
16
Cover image for ArcTrieval. Medical Correspondence Management System

.NET
ArcTrieval. Medical Correspondence Management System .NET Framework , TinyMCE , PayPal Integration , Google Maps API , SMTP , ASP.NET (http://ASP.NET)- HIPAA-compliant, a web-based medical correspondence management system that allows: - View tables of queries containing large volumes of data - Dynamically edit mail headings and synchronize between a mailbox and Arctrieval account - Generate PDF reports on the fly - Automatically apply localization settings based on user preferences from the browser - Process received fax messages and synchronize with Arctrieval system
1
80
Cover image for LimpidMarkets. Precious Metals Platfrom
Web DevelopmentThe
LimpidMarkets. Precious Metals Platfrom Web DevelopmentThe platform for London precious metals markets. LimpidMarkets has such clients as HSBC, BNP Paribas, JP Morgan, Morgan & Stanley, Deutsche Bank, UniCredit Bank, and more.
1
84
Cover image for ParamountDAX. Cryptocurrency Exchange Platform

Paramount DAX
ParamountDAX. Cryptocurrency Exchange Platform Paramount DAX partners with BitGo, 4STOP, Genesys, INDACOIN, CoinPaprika, CoinCodex, and more.
1
85