Darya Petrashka
Senior Data Scientist at SLB | AWS Community Builder
Szczytno, Poland
Actions
AWS Community Builder, works as a Senior Data Scientist at SLB. She is passionate about data and its usage for problem-solving. The area of interest includes NLP, GenAI, as well as working with AWS services. An eternal student, she likes taking part in online schools, courses, and workshops.
Badges
Area of Expertise
Topics
RAG Without Breaking the Bank: Create a Bedrock Agent with S3 Vector-Powered Knowledge Bases
Building RAG-powered agents doesn't have to come with sky-high vector database costs. In this session, learn how to create an Amazon Bedrock Agent backed by Amazon S3 Vectors, a new low-cost vector storage option that cuts vector storage and query costs by up to 90%.
We'll walk through:
1. Setting up a Bedrock Knowledge Base using S3 Vectors for semantic retrieval
2. Creating a Bedrock Agent that can reason over a large-scale knowledge base
3. Designing embeddings, chunking, and metadata filtering strategies for accurate results
Whether building internal copilots, customer support bots, or document search, this talk shows how to do RAG at scale without the price tag of high-performance vector databases.
This is ideal for developers and teams looking to build practical, budget-friendly GenAI apps on AWS.
Building a Cost-Effective Research Assistant with Bedrock Agents and S3 Vectors
Building RAG-powered agents doesn’t have to come with sky-high vector database costs. In this session, learn how to create an
Amazon Bedrock Agent backed by Amazon S3 Vectors, a new low-cost vector storage option designed to analyze research
papers efficiently and affordably. We’ll walk through:
1. Using a custom parsing workflow with AWS Lambda to process research papers
2. Setting up a Bedrock Knowledge Base using S3 Vectors for scalable semantic retrieval
3. Building a Bedrock Agent that can reason over parsed research content
4. Managing retrieval permissions with metadata filtering for fine-grained access control
Whether you’re building research assistants, internal copilots, or large document-analysis pipelines, this talk shows how to
extract insights at scale without the cost of traditional vector databases.
By the end of this talk, attendees will understand how to design a low-cost retrieval pipeline, know how to combine Bedrock
Agents with S3 Vectors effectively, and be equipped to build their own research analysis assistant on AWS.
Building a Cost-Effective E-Commerce Assistant with LangChain Agents and S3 Vectors
Creating intelligent assistants that query large datasets doesn’t have to be expensive. In this session, you’ll learn how to build a LangChain-powered agent that queries an Amazon S3 Vectors store built from CSV product review data, providing scalable, low-cost semantic search for e-commerce insights.
We’ll cover:
1. Vectorizing content for semantic search using S3 Vectors as the low-cost vector database.
2. Building a LangChain agent backed by OpenAI models to reason over query results.
3. Implementing role-based access control with metadata filtering, ensuring users see only what they’re allowed to access.
4. Query decomposition and rephrasing to improve retrieval accuracy and relevance.
By the end of the session, attendees will understand how to design a cost-efficient retrieval pipeline, combine LangChain agents with S3 vectors, and implement secure, relevant search over structured CSV data.
"You are an intelligent business analyst": how i learned to talk to business
I never planned to become a business analyst. In fact, I avoided it. I imagined endless meetings, unclear requirements, and conversations that had nothing to do with “real” technical work. I wanted to stay hands-on as a developer and data scientist.
But reality proved something important: you can't escape the business side if you want to build meaningful solutions. And once I learned how to talk to business stakeholders, everything changed: my impact, my influence, and the outcomes of the projects I worked on.
In this talk, we’ll explore the practical business skills every developer needs but is rarely taught:
• How to identify key stakeholders and understand what they really want
• How to navigate communication in international, cross-functional teams
• How to uncover business pain points before they become blockers
• How to fix broken communication loops
• How to become the go-to technical partner the business trusts
By the end, you won’t just see yourself as a strong technical contributor—you’ll see how to position yourself as an essential part of the broader business ecosystem, shaping better decisions and delivering solutions that truly matter.
Building a Bedrock agent for document processing
This session provides a step-by-step guide to building a Bedrock agent for document processing, showcasing how to automate tasks such as extracting key information, querying and altering documents, summarizing content, and organizing data. Attendees will learn how Bedrock agents work, explore potential use cases, and understand the concept of action groups. The session also dives into knowledge bases, covering their purpose, use cases, creation, querying, and integration with Bedrock agents. With a live demonstration and insights into integrating AWS services like Lambda, Textract, DynamoDB, and S3, this session equips attendees to build intelligent and efficient document workflows.
You don’t think about your Streamlit app optimization until you deploy it to AWS
Building Streamlit apps is easy for Data Scientists - but when it’s time to deploy them to the cloud, challenges like slow model loading, scalability, and security can become major hurdles. This talk bridges two perspectives: the Data Scientist who builds the app and the MLOps engineer who deploys it. We'll dive into optimizing model loading from Hugging Face Hub, implementing features like autoscaling and authentication, and securing your app against potential threats. By the end of this talk, you’ll be ready to design Streamlit apps that are functional and deployment-ready for the AWS.
Empower your Bedrock agent with GraphRAG
While traditional RAG excels in searching across unstructured documents, it often struggles to grasp intricate relationships between entities. GraphRAG overcomes this limitation by integrating knowledge graphs, enabling your solutions to understand and utilize these complex relationships.
In this session, we’ll explore how to supercharge Bedrock agents with GraphRAG, leveraging Amazon Neptune as the graph database alongside the Bedrock knowledge base. As part of the last re:Invent announcements, we'll see how this approach delivers precise, contextually rich answers.
The session centers on a use case in the tourism industry, but the techniques presented extend seamlessly across diverse domains such as document processing, healthcare, finance, media, and beyond.
Building an AI Chat Assistant with Amazon Bedrock Agent
During the session, you will discover the potential of AWS's groundbreaking announcement from last year—the Bedrock agent. Focused on building a smart assistant, this session will provide insights into how the Bedrock agent leverages the reasoning capability of foundation models (FMs) to support workflow orchestration and automation.
Explore the capabilities of the Bedrock agent through a captivating example project, where you will witness its ability to monitor updates on messaging platforms like Telegram, seamlessly translate messages, summarize content, and extract actionable insights.
By the end of this 30-minute talk, every attendee will gain a deeper understanding of the possibilities that the Bedrock agent holds for optimizing productivity and facilitating smart assistance. Join me as we embark on this exploration of the Bedrock agent's potential.
NLP on AWS: SageMaker and high-level services
NLP (natural language processing) nowadays is a very popular component of machine learning. AWS proposes a huge variety of tools to solve various NLP tasks.
AWS SageMaker which enables data scientists to quickly and easily train and deploy machine learning models has built-in NLP algorithms. The BlazingText algorithm (text classification) and 2 topic modeling algorithms (LDA and Neural Topic Model) will be covered during the talk.
However, to solve a specific NLP task, it is not always necessary to write any code, AWS offers many high-level services: Amazon Translate can translate many languages, Amazon Transcribe performs speech-to-text tasks, Amazon Polly converts text into speech, and Amazon Lex builds chat-bots. There is also Amazon Textract that performs document text detection and analysis tasks. All the mentioned services can be used in different domains like retail, commerce, education, medicine, etc.
An example project of a voicing chatbot combining Lex, Polly, and Lambda will be provided.
AWS Community Day Slovakia User group Sessionize Event Upcoming
Øredev 2025 Sessionize Event
AWS Community Day Baltic Sessionize Event
AWS Community pre:Invent Warmup Sessionize Event
AWS Community Day Italy 2025 Sessionize Event
AWS Community Day - Hungary 2024 Sessionize Event
AWS Community Day Nordics 2024 Sessionize Event
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top