Radu is available for hire
Hire RaduRadu Marian
Verified Expert in Engineering
Data Science Expert and Developer
Indian Trail, NC, United States
Toptal member since December 31, 2024
Bio
Radu is a senior GenAI and knowledge graph solution architect delivering solutions for cybersecurity, investments, agriculture, and economy domains. For over 25 years, he has accomplished numerous proofs of concept and pilot projects, led standardization efforts at OASIS and ITU-T, and was granted patents for inventions in cybersecurity and enterprise knowledge management. Radu holds a master's degree in computer science from UNC Charlotte.
Portfolio
OpenControls
Generative Artificial Intelligence (GenAI), Large Language Models (LLMs)...
Fidelity Investments
Neo4j, Snowflake, Python, Cypher, Plotly, LangChain, OpenAI, Microsoft Power BI...
Gro Intelligence
RDF, Stardog, Python, Jupyter, Slack, GitHub, Domain Ontology Modeling...
Experience
- Python - 12 years
- Knowledge Graphs - 12 years
- Natural Language Processing (NLP) - 10 years
- RDF - 10 years
- Neo4j - 6 years
- LangChain - 2 years
- OpenAI - 2 years
- Generative Artificial Intelligence (GenAI) - 1 year
Availability
Part-time
Preferred Environment
Neo4j, Generative Artificial Intelligence (GenAI), Python, RDF, Knowledge Graphs, SQL, LangChain, Machine Learning, GRC, Cyber Threat Intelligence (CTI)
The most amazing...
...thing I am developing—OpenCaply—a standard-based GenAI GRC agentic ecosystem that reduces GRC cost and risk by embedding AI agents into GRC operation teams.
Work Experience
GenAI and Knowledge Graph Architect
2024 - PRESENT
OpenControls
- Continued to develop OpenControls and OpenCaply, a standard-based GenAI GRC agentic ecosystem that reduces GRC cost and risk by embedding AI agents into GRC operation teams.
- Kept advancing OpenControls OASIS draft standard and knowledge graph model to commoditize 50% of enterprise GRC efforts.
- Continued to develop OpenControls knowledge graph reference implementation as a Stardog Knowledge Kit by integrating NIST AI RMF Playbook, MITRE Atlas, and OWASP Top 10 for LLM app compliance and threat response.
- Mapped enterprise AI controls to MITRE ATT&CK, NIST CVEs, and NIST CSF, enabling actionable compliance governance and risk management.
- Designed a GRC Vendor Product Features Mapping Assistant using LangChain and LangGraph to generate OpenControls Capabilities Taxonomy with suggested mapping to various GRC vendor product features vetted by human experts.
Technologies: Generative Artificial Intelligence (GenAI), Large Language Models (LLMs), LangChain, LangGraph, Neo4j, Stardog, ChromaDB, Python, Retrieval-augmented Generation (RAG), GraphRAG, Artificial Intelligence (AI), Neo4j Graph Data Science (GDS), Feature Engineering, PyTorch, Technical Leadership, Architecture, Enterprise Architecture, ChatGPT, Knowledge Graph Modeling, Vector Databases, Prompt Engineering, Full-stack, OpenAI API
Senior Knowledge Graph Architect | GenAI Data Scientist
2023 - 2024
Fidelity Investments
- Architected a financial advisor knowledge graph solution by integrating ETL from Snowflake to Neo4j and provided actionable business insights derived from exploratory analytics on production data.
- Performed advanced graph data analysis using Cypher and Python libraries, including Jupyter, Matplotlib, Seaborn, and Plotly, uncovering trends like advisors' most productive years and clustering top performers.
- Co-developed a graph data science pipeline to identify clusters of best-performing advisors using Degree Centrality, WCC, PageRank, Leiden, Louvain, and Node Similarity with embeddings.
- Created an advisor journey map knowledge graph to recommend the subsequent best actions, improving advisor engagement strategies.
- Piloted an interaction intent knowledge graph by extracting and standardizing intents from Salesforce Quarterly Business Review text via Azure LLM, LangChain, and LlamaIndex frameworks.
- Applied prompt engineering to align LLM-generated intents with a standardized taxonomy, enabling integration into the advisor journey map knowledge graph.
- Evaluated feature engineering techniques for customer lifetime value (CLTV) prediction.
- Developed a Power BI proof of concept (POC) to visualize Neo4j graph data using REST APIs and advanced visuals like Gantt and Sankey charts.
Technologies: Neo4j, Snowflake, Python, Cypher, Plotly, LangChain, OpenAI, Microsoft Power BI, Natural Language Processing (NLP), Knowledge Graphs, Data Science, Jupyter, Amazon Web Services (AWS), Large Language Models (LLMs), GraphRAG, PostgreSQL, Artificial Intelligence (AI), LlamaIndex, Neo4j Graph Data Science (GDS), Feature Engineering, Technical Leadership, Architecture, Enterprise Architecture, Knowledge Graph Modeling, Prompt Engineering, OpenAI API
Senior Ontologist Knowledge Graph Architect | Knowledge Engineer
2022 - 2023
Gro Intelligence
- Migrated CSV reference data and WebProtege ontology files to RDF repositories, including Stardog, Ontotext GraphDB, and AWS Neptune, for domains including agriculture, economy, and units of measure.
- Developed Jupyter notebooks for editing knowledge graphs using CSV inputs, with outputs stored in GitHub and RDF repositories.
- Initiated a knowledge editing solution by integrating Slack, Google Sheets, and RDF repositories using Python, streamlining knowledge graph updates.
- Visualized knowledge graphs in Tableau, providing clear insights and enhancing communication with stakeholders.
- Managed RDF repository operations, including evaluation, configuration, and maintenance for Stardog and Ontotext GraphDB repositories.
Technologies: RDF, Stardog, Python, Jupyter, Slack, GitHub, Domain Modeling, Domain Ontology Modeling, Plotly, Knowledge Graphs, SQL, OntoText GraphDB, Amazon, Amazon Web Services (AWS), PostgreSQL, Artificial Intelligence (AI), Technical Leadership, Architecture, Enterprise Architecture, ChatGPT, Knowledge Graph Modeling
Solution Architect | Research, Development, and Innovation Expert | Global Info Security Analyst
2007 - 2022
Bank of America
- Co-authored the OpenControls draft specification for mapping cybersecurity information, enabling governance, compliance, and operations via a reference knowledge graph.
- Presented the Cyber Control Ontology at the 2020 Knowledge Graph Conference, addressing cybersecurity governance and control vendor product mappings.
- Built a natural language processing (NLP) based framework to map regulatory documents like 14q and 14m to bank requirements using a Regulation Ontology and Lymba K-Extractor.
- Demonstrated mapping of business processes in enterprise taxonomy to application functionalities, enhancing enterprise asset governance via a knowledge graph.
- Proved TigerGraph's scalability for fraud detection, showing it to be four times faster at data loading and twice faster at querying than Neo4j for large datasets.
- Reduced manual fraud detection time from two hours to five minutes by transforming Splunk data into graph format and visualizing it with Linkurious for commercial banking.
- Automated contract parsing for vendor and cost extraction, reducing the time per document from 20 minutes to five minutes with Stardog and Linkurious visualization.
- Developed the Task-Based Access Management Standard (X.1257) approved by ITU-T in 2016, aligning IAM roles with business processes for accurate entitlement reviews.
- Built a POC for annotating threat reports with curated concepts from the MITRE ATT&CK framework to enhance the detection and prevention of adversary attacks.
Technologies: RDF, Stardog, Splunk, Neo4j, Linkurious, Python, Natural Language Processing (NLP), MITRE ATT&CK, Enterprise Cybersecurity, Enterprise Data Management (EDM), Plotly, Knowledge Graphs, Application Architecture, SQL, Jupyter, OntoText GraphDB, PostgreSQL, Neo4j Graph Data Science (GDS), Technical Leadership, Architecture, Enterprise Architecture, Knowledge Graph Modeling, Full-stack
SOA Enterprise Architect | Application Architect | Developer
2002 - 2007
ING Group
- Defined and specified SOA and enterprise service bus (ESB) requirements for the defined contribution line of business, fostering team adoption of SOA core principles such as document/literal and process composability.
- Influenced enterprise teams to align with SOA practices, improving consistency and efficiency across the organization by adopting scalable and reusable architectural standards.
- Designed and implemented SOA-based solutions, enabling seamless integration of business processes and ensuring adherence to industry best practices.
Technologies: Application Architecture, Java, Technical Leadership, Architecture, Enterprise Architecture
Lead Client-side Java Developer
1998 - 2000
IBM
- Headed the development of the GUI for the electronic parts catalog application, providing Navistar International with seamless access to truck parts data via the internet.
- Designed and implemented a client-side component framework using Java and Swing, ensuring reusable and scalable interface components.
- Collaborated with a team of eight to deliver an efficient intranet application that improved parts accessibility and streamlined operations for the client.
Technologies: Java, Technical Leadership
Experience
TrainSim | A Gamified GenAI Training and Customer Simulation Assistant
At the NODES 24 conference, I presented the TrainSim Assistant prototype that extracts relevant information from a PDF of a restaurant menu, generates hypothetical questions, and loads into a Neo4j graph database for subsequent GraphRAG interaction. This solution leveraged Jupyter for ETL purposes, LangChain for interacting with OpenAI and Neo4j, and Streamlit for delivering a gamified training simulation MVP user experience. The presentation slides are available at TrainSim-nodes24-presentation.pptx.
OpenControls | An OASIS Draft Standard and Knowledge Graph Reference Implementation
I am developing the OpenControls knowledge graph model and reference implementation. OpenControls is a draft OASIS standardization effort that maps MITRE ATT&CK and MITRE D3FEND, NIST CVEs, NIST CPEs, NIST Special Publication 800-53, and the NIST Cybersecurity Framework to enterprise assets, controls, and capabilities for compliance and risk governance purposes.
NIST AI RMF playbook, MITRE ATLAS, and OWASP Top 10 for LLM apps were added to the OpenControls knowledge graph to map and prove enterprise AI controls compliance and provide just-in-time recommendations on current and emerging AI threats.
NIST AI RMF playbook, MITRE ATLAS, and OWASP Top 10 for LLM apps were added to the OpenControls knowledge graph to map and prove enterprise AI controls compliance and provide just-in-time recommendations on current and emerging AI threats.
OpenCaply | A Standards-based GenAI GRC Agentic Ecosystem
OpenCaply is a standard-based GenAI GRC agentic ecosystem that reduces GRC cost and risk by embedding AI agents into GRC operation teams. OpenCaply leverages the OASIS draft standard OpenControls knowledge graph to automate and/or recommend just-in-time GRC operation actions to the GRC team, thereby reducing the cost of human effort and cybersecurity risk.
Interaction Intent Knowledge Graph Generation Pilot Using GraphRAG GenAI Approach
Implemented a POC and pilot that extracted interaction intents from Quarterly Business Review text stored and forgotten in Salesforce using Cloud Azure LLM via LangChain and LlamaIndex frameworks. These interaction intents were generated by LLM and vetted by business analysts.
Carefully crafted LLM prompt engineering helped suggest new or existing parents' mappings to a standardized interaction intent taxonomy, with the final goal of generating value-added investment advisor journey map knowledge graph nodes and relationships.
Carefully crafted LLM prompt engineering helped suggest new or existing parents' mappings to a standardized interaction intent taxonomy, with the final goal of generating value-added investment advisor journey map knowledge graph nodes and relationships.
Education
1996 - 1997
Master's Degree in Computer Science
University of North Carolina at Charlotte (UNC Charlotte) - Charlotte, NC, USA
Certifications
JULY 2023 - PRESENT
Unsupervised Learning, Recommenders, Reinforcement Learning
DeepLearning.AI & Stanford University | via Coursera
JULY 2023 - PRESENT
Advanced Learning Algorithms
DeepLearning.AI & Stanford University | via Coursera
JULY 2023 - PRESENT
Supervised Machine Learning: Regression and Classification
DeepLearning.AI & Stanford University | via Coursera
MARCH 2023 - PRESENT
AWS Certified Cloud Practitioner
Amazon Web Services
FEBRUARY 2023 - PRESENT
Neo4j Certified Professional Certification
Neo4
Skills
Libraries/APIs
OpenAI API, XGBoost, TensorFlow, PyTorch
Tools
Jupyter, Plotly, Microsoft Power BI, Slack, GitHub, Splunk, GraphRAG, ChatGPT
Languages
Python, Cypher, RDF, Java, Snowflake, SQL
Paradigms
Application Architecture, Anomaly Detection
Storage
Neo4j, PostgreSQL
Platforms
Stardog, Amazon, Amazon Web Services (AWS), AWS Lambda
Frameworks
LangGraph, LlamaIndex, Streamlit
Other
Domain Ontology Modeling, Enterprise Cybersecurity, Knowledge Graphs, Domain Modeling, Technical Leadership, Architecture, Enterprise Architecture, Knowledge Graph Modeling, LangChain, OpenAI, OntoText GraphDB, Linkurious, Natural Language Processing (NLP), MITRE ATT&CK, Enterprise Data Management (EDM), Generative Artificial Intelligence (GenAI), Data Science, Large Language Models (LLMs), Retrieval-augmented Generation (RAG), Artificial Intelligence (AI), Neo4j Graph Data Science (GDS), Feature Engineering, Prompt Engineering, Full-stack, Machine Learning, GRC, Cyber Threat Intelligence (CTI), Logistic Regression, Artificial Neural Networks (ANN), Linear Regression, Decision Trees, Recommendation Systems, Model Development, Tree Ensembles, Unsupervised Learning, Reinforcement Learning, Collaborative Filtering, ChromaDB, Threat Intelligence, AI Algorithms, Vector Databases
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