Gregory Kuklin

• • Designed and integrated a FastAPI-based RAG pipeline into the company's core monolith application for internal document search.

• • Implemented vector database storage, making data retrieval about 40% faster for the team.

• • Built LLM-driven workflows within the application to automatically route and classify internal company documents.

• • Achieved a 94% accuracy rate in extracting key entities, which cut down manual customer support tasks by 30%.

• • Set up strict JSON/structured output validation layers and custom prompt chains, keeping LLM hallucination rates in production below 1%.

• • Maintained and expanded backend functionality, dockerized the application components, set up automated CI/CD pipelines, and managed deployments to AWS.

• • Built an end-to-end processing tool using Tesseract OCR and PyTorch to read and parse scanned business documents (like invoices and contracts), validating extracted data for consistency.

• • Developed and tuned REST APIs using FastAPI and PostgreSQL. Speeded up database performance by 25% by optimizing SQL queries and adding a Redis caching layer.

• • Handled projects from scratch from early technical design and writing code to setting up basic monitoring after deployment.