Roman is available for hire
Hire RomanRoman Bartusiak
Verified Expert in Engineering
Artificial Intelligence Developer
Wrocław, Poland
Toptal member since March 26, 2018
Bio
Roman builds production AI and data-intensive systems across NLP, semantic search, knowledge graphs, agentic workflows, and back-end platforms. With years of engineering experience, he brings staff-level ownership, CTO/founder execution, 9 publications, and 8 R&D projects. He turns research into reliable, measurable, deployable software, with a strong focus on architecture, performance, evaluation, and mentoring.
Portfolio
Panasonic
Large Language Models (LLMs), Agentic AI, Web Architecture, React, APIs, Kotlin...
Causality
Machine Learning, Natural Language Processing (NLP), Semantic Search, Word2Vec...
Grantland Ltd.
Artificial Intelligence (AI), AI Consulting, Deep Neural Networks (DNNs), APIs
Experience
- Microservices - 8 years
- Deep Neural Networks (DNNs) - 7 years
- Python - 7 years
- Distributed Computing - 7 years
- Kotlin - 6 years
- Artificial Intelligence (AI) - 5 years
- Large Language Models (LLMs) - 5 years
- React - 4 years
Preferred Environment
IntelliJ IDEA, Linux, Kotlin, Python, Artificial Intelligence (AI), Scala, Java, Machine Learning, Kubernetes, Amazon Web Services (AWS)
The most amazing...
...thing I've coded is an open-source library, SparklingGraph, that helps process graphs in distributed environments.
Work Experience
Staff Machine Learning Engineer, Global Applied AI
2026 - PRESENT
Panasonic
- Architected and built KanpAI, Panasonic's enterprise agentic AI platform, from the ground up.
- Created the core Agent SDK, defining the agent packaging model, runtime contracts, and secure deployment flow for BYOA/BYOM use cases.
- Designed back-end services for agent deployment, orchestration, and lifecycle management across multi-tenant cloud, on-prem, and edge environments.
- Helped scale KanpAI to 20+ agents across 10+ domains and 20+ model families, supporting 70+ concurrent enterprise users.
Technologies: Large Language Models (LLMs), Agentic AI, Web Architecture, React, APIs, Kotlin, Azure, Cloud, Docker Compose, GitHub, Visual Studio Code (VS Code)
Co-founder
2021 - PRESENT
Causality
- Co-founded Causality, an applied-AI studio delivering research-grade software for clients across machine learning and NLP.
- Built semantic search and structured-data extraction products for client engagements, from research prototype to production deployment.
- Led bespoke AI engineering work, including EU-funded R&D projects, deploying solutions both cloud-based and fully on-premise.
Technologies: Machine Learning, Natural Language Processing (NLP), Semantic Search, Word2Vec, PyTorch, Deep Neural Networks (DNNs), React, Distributed Computing, APIs
Independent Consultant
2020 - PRESENT
Grantland Ltd.
- Advised Grantland's clients on AI and technology components of EU-funded development projects.
- Scoped and implemented AI-driven solutions for businesses and institutions within grant-funded R&D and innovation initiatives.
- Guided projects from technical feasibility assessment through to execution for grant-funded innovation programs.
Technologies: Artificial Intelligence (AI), AI Consulting, Deep Neural Networks (DNNs), APIs
Owner (Freelance)
2016 - PRESENT
Roosters IO
- Implemented person detection and tracking for both wifi and Bluetooth using Rust.
- Developed a method to estimate a person's gender and age using Kairos SDK.
- Enabled people detection in a CCTV stream using multiple machine-learning methods.
- Created web applications using Rust Rocket and React.
- Determined a person's gender and estimated the age using Affectiva SDK.
- Built an RFID tag scanner using Rust and User Diagram Protocol communication.
Technologies: Big Data, Akka Actors, Artificial Intelligence (AI), Apache Spark, JavaScript, Rust, Python, Scala, Java, Unix, RDBMS, Arch Linux, MongoDB, Python 3, Deep Neural Networks (DNNs), React, APIs, jQuery, Sass, Compass, Responsive Web Apps, C++, Android
Senior Machine Learning Engineer, Panasonic Well
2024 - 2026
Panasonic
- Designed and implemented core ML/AI platform services, applying software-engineering rigor (testing, monitoring, CI/CD) to research-driven components.
- Took solutions from idea through design, technology selection, deployment, and monitoring, owning them end-to-end.
- Profiled and optimized existing models and pipelines for performance and accuracy, converting findings into concrete engineering improvements.
- Established platform practices and technical standards that carried forward into Panasonic's Global Applied AI agentic platform (KanpAI).
Technologies: Python, Large Language Models (LLMs), gRPC, Machine Learning, Kotlin, Kubernetes, API Development, Containers, Containerization, LangChain, FastAPI, Claude, Claude Code, Workflow, Automation, Web Architecture, Deep Neural Networks (DNNs), React, APIs, DevOps, GitLab CI/CD, Jenkins, Travis CI, Git, GitLab, OOP Designs
Scala Expert
2020 - 2024
Asia Wealth Platform Pte Ltd
- Contributed to a highly scalable platform using Akka, Lagom, and Kubernetes.
- Modified tools to improve version upgrades, including implementing breaking changes on Lagom.
- Migrated to a Kotlin and Micronaut microservices-based trader platform.
- Redesigned investment processing pipelines serving hundreds of thousands of customers, cutting end-to-end run time by 2x.
- Optimized portfolio rebalancing algorithms and batch execution for 2x faster runs while preserving correctness, auditability, and operational controls.
- Owned staff-level architecture for JVM-based investment services using Lagom, Scala, Kubernetes, Helm, and CQRS+ES patterns.
- Strengthened test coverage, observability, and rollback safety across critical fintech workflows, mentoring engineers on performance and data-model reviews.
Technologies: Scala, Akka, Lagom, Cassandra, Kotlin, Micronaut, Apache Kafka, Node.js, NoSQL, REST APIs, SQL, TypeScript, Unix, RDBMS, Temporal, Dashboards, Datadog, API Architecture, APIs, Testing, Clean Code, Mockito, Web Architecture, Distributed Computing, SBT, Akka Streams, Play 2, PostgreSQL, Redis, Elasticsearch, Sangria, AssertJ, Test-driven Development (TDD), Functional Programming, MySQL, Spring 5, Sagas
CTO | Co-founder
2017 - 2020
IamBot
- Built image similarity models using TensorFlow. Developed convolution DDNs for image representation and cross-image similarity calculation.
- Developed the back end using the Play framework, Akka, and Cats.
- Implemented Akka-based work distribution. Sharded work in order to increase the scalability and parallelism of the app.
- Developed a React, Redux, Sagas, and Relay dashboard using the Metronic template.
- Worked with natural language processing (NLP) and natural language understanding (NLU) based on word vectors.
- Utilized GoogleML in order to train machine learning models.
- Designed and implemented deployment procedures using Kubernetes and Helm.
- Utilized multiple languages in microservices that communicated using gRPC.
- Created transformer-based language models using PyTorch.
Technologies: Akka Actors, Convolutional Neural Networks (CNNs), Neural Networks, Artificial Intelligence (AI), Machine Learning, gRPC, Helm, Kubernetes, Relay, GraphQL, Rust, Python, React, TensorFlow, Play Framework, Scala, Unix, RDBMS, Deep Neural Networks (DNNs), Distributed Computing, Redux, Redux Thunk, Git, GitFlow, Swagger, JavaScript 6
IT Consultant | Research Assistant
2015 - 2016
Wrocław University of Science and Technology
- Implemented an SSO platform based on OAuth 2 using Spring.
- Integrated multiple services using Docker. Deployed on custom infrastructure.
- Maintained experimental environments for heterogeneous clusters.
- Managed Spark clusters based on the BEM scientific cluster.
- Executed data backups using IBM Tivoli to guarantee data retention.
Technologies: Angular, Spring, JavaScript, Java, Unix, Apache Maven, EMR, SSH, Bash, LaTeX, Pair Programming, REST, JSON, GraphX, Python 2, DRY
Software Architect
2013 - 2015
Nokia Networks
- Developed Vaadin portlets and components (GWT, JavaScript).
- Created JEE web services deployed on Tomcat and JBoss.
- Built a proof of concept (POC) using Angular to interact with existing software.
- Created a full deployment solution using Jenkins and Docker.
Technologies: Angular, GWT, JEE, Vaadin, Java, Unix, SOAP, Scrum, Zsh, Oh My Zsh
Web Developer
2012 - 2013
Chop-Chop
- Supported a Magento store development services (custom themes, plugins, features).
- Wrote Java software that parsed PSD files to optimize PSD2HTML.
- Translated PSD files to HTML code using custom software.
Technologies: JavaScript, CSS3, HTML5, Unix, Adobe Photoshop, GitHub Pages, Sublime Text, Web Development, Redmine
Experience
BORSUK - Blob-oriented Retrieval with Segmental Unified KNN
https://causality.pl/borsuk/BORSUK keeps the entire index as immutable Parquet objects in the S3, MinIO, SeaweedFS, GCS, or Azure storage you already pay for — and answers a query with a few hundred bytes of resident memory. No always-on RAM cluster to provision, scale, or feed between queries.
SFORA - Deep Metric Learning Library
https://causality.pl/sfora/SFORA is a deep metric learning library. Its method, HERD — Hypergraph EMA-teacher Relational Distillation — plus a multi-model ensemble beat the reported SOTA on the same architecture for CUB-200 by more than 1 point.
DGX Spark Unsloth
https://github.com/riomus/dgx-spark-unslothBuilt a toolkit for running Unsloth workloads on NVIDIA DGX Spark at full performance by leveraging Spark-optimized official NVIDIA Docker images and developing benchmarking and evaluation methods to validate system performance. Enabled users to fully utilize the hardware's capabilities and maximize throughput.
SparklingGraph
https://sparkling-graph.github.io/Using SparklingGraph, you can do fast computations on your graph data. You can leverage the computational power of multiple nodes and also run computations on your own laptop, thanks to Apache Spark. SparklingGraph provides many graph measures, but also gives you the ability to easily load graphs from many common formats like CSV, GML, and more.
Sentinent Analysis for Polish Language using SVM and BoW
https://github.com/riomus/polish-sentimentA self-contained project that enables users to classify texts in terms of sentiment. Using a provided Docker image, you can easily use state-of-the-art NLP methods for the Polish language and achieve cutting-edge results in sentiment analysis tasks. The provided API classifies text into two sentiment classes: positive and negative.
HMM.js
https://github.com/riomus/HMM.jsThese are Hidden Markov models implemented in JavaScript with Viterbi algorithms for learning. The library gives you the ability to create and train Hidden Markov Models that can be used in many tasks like sequence classification, NLP, and more. By using low-level JavaScript primitives, the library achieves low classification and training times.
Gesture Recognition
https://github.com/riomus/GestureRecognitionThis is a program for gesture recognition for JavaScript using Hidden Markov Models. The project uses HMM.js and OpticalFlow.js to classify user gestures captured by a webcam. Using that library, you can create games that utilize gestures as a controller.
Vibe.js
https://github.com/riomus/vibe.jsA ViBE background subtraction algorithm implemented in JavaScript (with performance optimizations). The library uses low-level JavaScript primitives to achieve low processing time per video frame. Using this library, you can create projects that can easily detect non-backgrounds on webcam video streams (or other video sources).
OpticalFlow.js
https://github.com/riomus/OpticalFlow.jsAn optical flow that was implemented in JavaScript. The library implements optical flow-based movement detection in video streams. It can be used to detect movements and process them to detect gestures. Using this library, you can detect the direction of the movement on each pixel between two video frames.
Education
2015 - 2021
PhD in Computer Science (Artificial Intelligence)
Wrocław University of Science and Technology - Wrocław, Poland
2014 - 2015
Master's Degree in Computer Science (Artificial Intelligence)
Wrocław University of Science and Technology - Wrocław, Poland
2010 - 2014
Engineer's Degree in Computer Science
Wrocław University of Science of Technology - Wrocław, Poland
Certifications
APRIL 2020 - PRESENT
Reactive Architecture (5): Distributed Messaging Patterns
Lightbend, Inc.
APRIL 2020 - PRESENT
Reactive Architecture (4): Building Scalable Systems
Lightbend, Inc.
APRIL 2020 - PRESENT
Reactive Architecture (3): Reactive Microservices
Lightbend, Inc.
MARCH 2020 - PRESENT
Reactive Architecture (2): Domain Driven Design
Lightbend, Inc.
MARCH 2020 - PRESENT
Reactive Architecture (1): Introduction to Reactive Systems
Lightbend, Inc.
OCTOBER 2015 - PRESENT
Machine Learning
Stanford University via Coursera
SEPTEMBER 2014 - PRESENT
Java Standard Edition 6 Programmer Certified Professional
Oracle University
APRIL 2014 - PRESENT
Functional Programming Principles in Scala
École Polytechnique Fédérale de Lausanne via Coursera
JULY 2013 - PRESENT
Introduction to Software Development on SAP HANA
SAP
Skills
Libraries/APIs
Node.js, GraphX, Akka Streams, REST APIs, Spark ML, React, Sangria, TensorFlow, AssertJ, PyTorch, vLLM, jQuery, API Development
Tools
Apache Maven, Helm, SBT, GitLab CI/CD, LaTeX, GitHub Pages, GitHub, GitLab, Git, Redux Thunk, Sublime Text, IntelliJ IDEA, Docker Compose, Oh My Zsh, Zsh, Jenkins, Travis CI, Redmine, Adobe Photoshop, Claude, Claude Code
Languages
Java, Python 2, Python 3, GraphQL, Scala, Python, Kotlin, HTML5, JavaScript, JavaScript 6, Rust, SQL, TypeScript, Bash, Sass, CSS3, C++
Frameworks
Apache Spark, Akka, Spark, Play Framework, Play 2, Relay, gRPC, Angular, Spring, Redux, Mockito, Spring 5, GWT, Swagger, Compass, Vaadin, Lagom, Micronaut
Paradigms
Microservices, Web Architecture, Distributed Computing, DevOps, Test-driven Development (TDD), Testing, Functional Programming, Scrum, Agile Software Development, Clean Code, API Architecture, Reactive Programming, DRY, REST, Pair Programming, Object-oriented Programming (OOP), Automation
Platforms
Docker, Arch Linux, Kubernetes, Amazon Web Services (AWS), Linux, Unix, Visual Studio Code (VS Code), JEE, Azure, Android, Apache Kafka, SAP HANA
Storage
MongoDB, Elasticsearch, MySQL, PostgreSQL, Redis, NoSQL, RDBMS, JSON, Cassandra, Datadog
Other
Akka Actors, Machine Learning, Natural Language Processing (NLP), Artificial Intelligence (AI), Neural Networks, Deep Neural Networks (DNNs), Convolutional Neural Networks (CNNs), Graphs, OOP Designs, APIs, Sagas, Generative Pre-trained Transformers (GPT), Large Language Models (LLMs), EMR, Responsive Web Apps, Web Development, Cloud, Big Data, Computer Vision, Hugging Face, Code Architecture, Message Queues, Scalability, Reactive Streams, Domain-driven Design (DDD), SOAP, GitFlow, SSH, Word2Vec, WebAssembly (Wasm), Temporal, Dashboards, Containers, Containerization, LangChain, FastAPI, Workflow, Semantic Search, AI Consulting, Agentic AI
Collaboration That Works
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
1
Share your needs
Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2
Choose your talent
Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3
Start your risk-free talent trial
Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.
Top talent is in high demand.
Start hiring