Maciej K.
Senior Backend Engineer & AI System Developer
Experience
Jan 2015 - Present
11 yearsSenior Backend Engineer & AI System Developer
- Designed and deployed more than 50 backend apps across analytics and AI platforms with over 90% test coverage.
- Architected hybrid search pipelines combining structured filters, BM25, and semantic vector similarity search.
- Built full scraping frameworks and automated pipeline orchestration for datasets exceeding 250 000 records across multiple verticals.
- Implemented retrieval-augmented generation architectures reducing manual workflow time by 40%.
- Integrated OpenAI and domain-specific models into production environments with secure async execution and observability.
- Twin4ever.com - AI assistant platform (Stack: FastAPI, Socket.IO, AsyncPG, OpenAI API (GPT-4/5), LangChain)
- Developed a real-time AI conversational platform enabling interactive and personalized dialogue experiences powered by OpenAI’s large language models.
- Built scalable WebSocket-based backends using FastAPI and Socket.IO, providing stateful, low-latency connections for thousands of concurrent users.
- Integrated OpenAI API for dynamic response generation, context retention, and multi-turn dialogue management.
- Implemented session-aware LLM pipelines leveraging LangChain to maintain conversational memory, execute function calls, and retrieve external knowledge dynamically.
- Engineered asynchronous processing with AsyncPG and Redis to queue and handle high-throughput conversational workloads efficiently.
- Designed extensible APIs for NLP-driven interactions such as summarization, sentiment analysis, and context classification.
- Trait.Tech - Blockchain data gateway (Stack: FastAPI, MongoDB)
- Developed secure gateway services for blockchain transaction data.
- Built resilient data-processing pipelines and administrative APIs with JWT authorization.
- AI-Powered Analytics Engine (Stack: FastAPI, PostgreSQL, Redis, GPT-4)
- Built event-driven analytics APIs handling 10 000 requests/min with sub-200ms latency.
- Created asynchronous pipelines for streaming analytics and LLM-based insight generation.
- integrated observability and alerting with Prometheus and Grafana metrics.
- Custom prefilled sticker display platform (Stack: FastAPI, Next.js, MongoDB, Docker, Google Cloud Run, Google Cloud Storage)
- Designed and implemented a scalable full-stack web platform enabling retailers to create and purchase custom, data-driven prefilled sticker displays with an interactive, visual ordering experience.
- Built a responsive Next.js frontend with smooth single-page transitions, real-time customization, and dynamic rendering of design previews.
- Developed a FastAPI backend with a MongoDB schema supporting flexible metadata structures for categories, product configurations, name drops, and order history.
- Implemented smart recommendation and weighting logic (80/20 mix of top-selling and new designs) driven by user context such as business type, location, and keywords.
- Created an automated SVG name-drop rendering pipeline that merges user-provided text into sticker templates and delivers printer-ready vector files directly to Google Cloud Storage.
- Deployed backend services on Google Cloud Run, enabling containerized scaling with zero-downtime updates and efficient resource utilization.
- Integrated secure business validation, account creation, and reorder functionalities, preserving user configurations through persistent MongoDB storage.
- Designed an admin-ready CMS structure enabling internal teams to update designs, categories, and configurations without developer intervention.
- JustAsk - AI shopping assistant (Role: Backend Architect and Golang Developer; Tech Stack: Golang, OpenAI API, pgvector, PostgreSQL, Docker)
- Built a four-layer Golang backend: scraping layer, normalization layer, vector storage layer, query layer.
- Ensured modular, testable components using TDD and continuous integration.
- Optimized vector queries for performance and relevance scoring.
- Reduced query latency by 60% and unified fragmented marketplace data into a single intelligent search system.
- Enterprise RAG Platform (Stack: LangChain, Qdrant, MongoDB, FastAPI, Hugging Face Transformers, spaCy, OpenAI API, Docker, Kubernetes)
- Architected and deployed a multi-tenant retrieval-augmented generation pipeline for enterprise-level contextual document search and knowledge retrieval.
- Designed and implemented an NLP processing pipeline for document ingestion, including tokenization, entity recognition, summarization, and embedding generation using transformer-based models.
- Automated chunking, metadata extraction, and semantic indexing, enabling highly efficient retrieval across large and diverse document corpora.
- Optimized embedding storage and similarity search using Qdrant with fine-tuned relevance scoring and hybrid search strategies.
- Integrated version-controlled prompts and NLP components through LangChain for consistent model orchestration across different tenants and data domains.
- Deployed scalable microservices to Docker and Kubernetes clusters with modular APIs supporting dynamic retrieval workflows and AI-driven contextual responses.
- Designed and maintained backend systems implementing clean architecture and SOLID principles.
- Achieved sub-200ms hybrid semantic search latency over datasets greater than 100 GB.
- Automated ingestion and normalization of multi-source scraping jobs with semantic indexing.
- Integrated AI modules for natural-language search, automatic classification, and agent orchestration.
- Maintained over 90% unit and integration test coverage throughout CI/CD pipelines.
Summary
Seasoned Backend Engineer and AI/ML Specialist with over 10 years of experience designing scalable backend systems, distributed microservices, and AI-driven products. Expert in Python as the primary development language, with additional proficiency in Golang and Next.js. Highly experienced in building web scraping frameworks, data pipelines, semantic search systems, and AI integrations powered by modern LLM and NLP technologies. Skilled in designing and deploying end-to-end NLP pipelines for text classification, summarization, embedding generation, and conversational AI. Adept at integrating intelligent APIs with React.js and Next.js frontends to deliver seamless, data-driven user experiences. Committed to clean architecture, modular design, and test-driven development (TDD), and experienced in deploying cloud-native, production-grade services for data processing and hybrid search.
Languages
English
AdvancedEducation
Oct 2011 - Jun 2015
Military University of Technology
Master of Science in Computer Science · Computer Science · Warsaw, Poland
Need a freelancer? Find your match in seconds.
Try FRATCH GPT More actions
Similar Freelancers
Discover other experts with similar qualifications and experience
- English
- Deutsch
2025 © FRATCH.IO GmbH. All rights reserved.