Patryk Orzechowski
Verified Expert in Engineering
Artificial Intelligence Developer
Havertown, PA, United States
Toptal member since June 2, 2026
Patryk is an AI solution architect and applied AI leader specializing in enterprise AI, intelligent automation, AI agents, and LLM systems. He combines deep expertise in ML, deep learning, NLP, and multimodal AI with a strong record of delivering scalable solutions that improve operational efficiency, accelerate decision-making, and create measurable business value. Patryk is the inventor of the EBIC algorithm and co-author of influential benchmark projects including PMLB, SRBench, and DIGEN.
Portfolio
Experience
- Machine Learning - 20 years
- Artificial Intelligence (AI) - 20 years
- Data Science - 20 years
- Benchmarking - 10 years
- AI Research - 10 years
- AI/ML Workloads - 8 years
- RAG Architecture - 5 years
- Large Language Models (LLMs) - 4 years
Preferred Environment
Artificial Intelligence (AI), Machine Learning, Data Science, Large Language Models (LLMs), Natural Language Processing (NLP), AI Architecture, RAG Architecture, AI/ML Workloads, Agentic AI, Model Context Protocol (MCP)
The most amazing...
...work I've done is building and executing a leading method for finding subsets in data (EBIC) and improving systematic literature review AI intelligence.
Work Experience
Staff Data Scientist, Life Sciences & AI
Intelligent Medical Objects
- Architected and deployed production AI intelligence services for a life sciences SaaS platform using Azure AI Foundry, AWS Bedrock, and LLM APIs.
- Built automated LLM evaluation and monitoring frameworks using LLM-as-a-judge methodologies and Claude Code to track reasoning quality, grounding, safety, and reliability across production AI systems.
- Shipped agentic AI features via MCP, LangChain, LangGraph, and Pydantic.
- Served as internal SME for LLM integration, AI governance, model validation, and responsible AI practices in healthcare environments.
Staff Data Scientist
Valo
- Designed novel ML pipelines for high-dimensional longitudinal EHR.
- Delivered AI-driven evidence-generation prototypes that accelerated research and improved analytical efficiency.
- Built automated AI/ML pipelines combining deep learning and classical ML.
- Improved internal tooling and scalable data infrastructure for clinical AI applications.
- Collaborated with cross-functional teams, key stakeholders, and clients.
Senior Data Scientist
University of Pennsylvania
- Developed a Python package with reproducible classification workflows with Docker containers.
- Implemented automated software testing with GitHub actions, improving existing pipelines.
- Led cloud infrastructure project (AWS/Python/GO), data ingestion project (Python), and retrieval-augmented generation (RAG) application.
- Created webhooks for cloud environments (AWS, Terraform, and Python).
- Analyzed clinical, genomic, and immunology datasets, deriving insight from multimodal data, graphs, and texts (NLP).
- Collaborated with physicians, immunologists, researchers, and technicians.
- Taught natural language processing (NLP) for UPenn's health course.
- Applied deep learning, graph neural networks (GNN), RAG, and large language models (LLM) to different problems.
- Performed ad-hoc analyses of next-generation sequencing (NGS), biological, and clinical data.
Postdoctoral Researcher in AI
University of Pennsylvania
- Designed and implemented a multi-GPU algorithm for subset detection EBIC—one of the leading biclustering methods in the field.
- Developed an R package for Bioconductor called runibic.
- Co-developed PMLB and srbench—influential data science benchmarks—each cited 500+ times (C++, Python, and Java).
Experience
Systematic Literature Review
https://www.imohealth.com/systematic-literature-review/Key capabilities included semantic query expansion using medical ontologies and controlled vocabularies, automated retrieval of scientific publications from biomedical databases, AI-assisted relevance screening, extraction of study characteristics and outcomes, and generation of structured evidence summaries. The platform incorporated human-in-the-loop validation workflows to maintain scientific rigor while significantly reducing manual review effort.
IMPACT
• Reduced the time required to conduct systematic literature reviews.
• Improved recall and precision of scientific literature searches through terminology-driven semantic search.
• Automated the extraction of key study attributes, outcomes, and evidence tables.
• Enabled researchers, medical affairs, and health economics teams to generate evidence more efficiently and at greater scale.
Multi-agent System for Drug Discovery
https://github.com/athril/agentic-target-evidenceEducation
Doctorate Degree in Computer Science
AGH University - Krakow, Poland
Master's Degree in Computer Science
AGH University - Krakow, Poland
Master's Degree in Automotive Engineering
AGH University - Krakow, Poland
Certifications
Hugging Face Agents Course
Hugging Face
Skills
Libraries/APIs
Hugging Face Transformers, OpenAI API, Claude API, Pydantic, Scikit-learn
Tools
ChatGPT, Claude Code, Claude, Seaborn
Languages
Python, SQL, C++, R, TypeScript, Java
Frameworks
LangGraph, Flask, Agentic Frameworks
Paradigms
Model Context Protocol (MCP), DevOps
Platforms
Docker, NVIDIA CUDA, Amazon Web Services (AWS), Google Cloud Platform (GCP), Langfuse
Industry Expertise
Bioinformatics
Storage
Graph Databases
Other
Artificial Intelligence (AI), Machine Learning, Data Science, Large Language Models (LLMs), Natural Language Processing (NLP), AI Architecture, RAG Architecture, AI/ML Workloads, Neural Networks, Data Classification, AI Research, R&D, Agentic AI, Multi-agent Systems, Benchmarking, MLflow, Data Engineering, Workflow Automation, AI Agents, Generative Artificial Intelligence (GenAI), API Integration, Prompt Engineering, Large Language Model Operations (LLMOps), Architecture, AI Design, AI Programming, Software Deployment, Scalable Platforms, Machine Learning Operations (MLOps), Agentic AI Systems, LangChain, OpenAI, Knowledge Graphs, RAG Pipelines, Retrieval-augmented Generation (RAG), Clinical Research, GPU Computing, Agentic Coding, Solution Architecture, LLM Reasoning, Agentic RAG Systems, Bioconductor, Regression, Amazon Bedrock AgentCore, Ontologies, Data Extraction, Anthropic, Data Curation, Ground Truth, Industry Benchmarking, Model Evaluation, Pipelines, AI Pipeline, ML Pipelines, Health Economics & Outcomes Research (HEOR), Hypothesis Testing, Drug Development, Biostatistics, Epidemiology, Electronic Health Records (EHR), Deep Learning, FastAPI
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
Share your needs
Choose your talent
Start your risk-free talent trial
Top talent is in high demand.
Start hiring