Srdjan Marinovic
srdjan dot marinovic at gmail dot com
I'm a computer scientist working on distributed graph platforms and analytics, and AI systems (both
symbolic and deep learning). I focus on
building secure, reliable, and scalable systems that bridge the gap
between cutting-edge research and practical engineering solutions.
My journey spans academic research at ETH Zurich and Imperial College London, co-founding
and scaling a graph analytics startup (acquired by PwC), and leading
enterprise-scale data and AI initiatives within PwC. I serve as an ML/data advisor to
Futurae on fraud-detection
In my leadership roles, I focus
on building and mentoring high-performing engineering teams thorugh
continuous learning and knowledge sharing.
Current Work at PwC
As a Tech-and-Innovation Director, I lead the development of
enterprise-scale AI and cloud solutions that integrate knowledge graphs,
machine learning, and federated data platforms. My approach emphasizes rapid
iteration and deployments, transforming complex technical challenges into
high-ROI business solutions.
Key Initiatives:
- Signal Graph "Operational AI" Platform - Advanced analytics platform that processes streaming time-series data using LLMs and knowledge graphs for context-aware operational intelligence. Our innovative symbolic post-processing layer ensures reliable AI-generated insights.
- Smart Venues Solution - Next-generation venue management systems that leverage IoT data and advanced analytics for enhanced operational efficiency.
- LLM ROI Monitoring in Software Development - Methodologies and tools to measure LLMs' reliability and effectiveness across development teams, focusing on the transition from PoCs to production-scale deployments. We also develop systems to enhance the effectivness of LLMs in the development cycle (both code generation and code reviews).
- Passive Organizational Network Analysis - Graph ML models on top of communication meta-data (such as email exchanges, chats, meetings, docuemnt collaborations) to understand informal organizational structures and influencers. Our product provids insights into how employment and team policies imapct team dynamics and organizational evolution. We also focus on tracking how junior team members are included and promoted within an organization.
SignalFrame Startup
As technical co-founder and CTO of SignalFrame (founded in
2015, acquired by PwC in 2021), I led a team of 20 engineers in building a
streaming temporal graph platform that processed data from ~1 billion IoT
devices monthly. Our daily stream consisted of ~100 million vertices with ~1
billion edges. On top of this knowledge graph we implemented advanced graph
embeddings for temporal and spatial analytics. Our products focused on
behavioral anomaly detection and structural forecasting.
Technical Innovations:
- Temporal graph analytics engine
- Streaming framework with feedback loops (Golang)
- Distributed graph query system (Golang)
- Graph-pattern classifiers (Python and Scala)
Research Background
- Database Security: Developed probabilistic attack prevention for SQL databases using Bayesian networks
- Distributed Systems Monitoring: Created metric temporal logics for handling compromised and incomplete system logs (Google collaboration)
- Decentralized Authorization: Built formal verification tools for complex authorization policies and authentication systems (Kaba Security collaboration)
- Developed soft symbolic AI models using multi-valued and probabilistic logics
- Created adaptive security models for decision-making under system failures
Selected Writings
Full publication list: DBLP
Patents: Google Patents