Kristina

Designed and optimized NLP pipelines for scalability and reduced inference latency by implementing batching, Kubernetes-based scaling, and experimenting with Python multiprocessing. Built modular microservices with FastAPI, containerized with Docker, and deployed on Kubernetes. Integrated PostgreSQL with optimized schemas and indexing to support high-throughput NLP workloads.

Developed an AI service enabling automated information extraction from financial documents, combining OCR preprocessing (Pytesseract) with transformer-based NLP models. Built retrieval-augmented generation (RAG) pipelines using HuggingFace embeddings and Milvus vector DB, experimenting with LangChain to orchestrate document ingestion, retrieval, and response generation. Applied prompt engineering to improve accuracy, with real-world validation against production financial datasets. Served the system via FastAPI, leveraging SQLAlchemy for ORM database management. Ensured reproducibility and maintainability by applying CI/CD, version control, and modular microservices architecture.

Led a cloud-hosted platform helping farmers monitor biodiversity via audio/image recognition, integrating TensorFlow models into production APIs. Deployed solutions on AWS using Docker containers, with data persisted in PostgreSQL and orchestrated workflows. Coordinated project scope through presales calls, client workshops, and estimation phases. Delivered demos and technical presentations to audiences of 100+ stakeholders. Managed a cross-functional team, ensuring code quality, structured testing, and knowledge-sharing practices.