Advisor · VP of AI Strategy
Svetlana Kharlamova
Svetlana Kharlamova, Ph.D., builds AI systems that work in production.
Her technical knowledge spans statistical and econometric modeling, forecasting and optimization, reinforcement learning, Edge AI, computer vision, neural networks, and NLP from Word2Vec and LSTM through BERT, GPT, and privately deployed LLMs.
She is VP of Engineering & AI on an adaptive healthcare platform, leading a 35-person software and data engineering organization against a $19M budget under HIPAA and GDPR. She architected its Digital Integration Hub with distributed YugabyteDB, Kafka and Flink streaming pipelines, a data lake, knowledge graphs, and Agentic AI services, to make previously siloed clinical and operational systems interoperable, and built a federated data network using federated learning inside Trusted Execution Environments (SGX/TDX, OCI confidential VMs) so models train on patient data that never leaves its vault. She owns attestation, enclave key management, and consent enforcement, and cut inference and compute spend 21% through FinOps discipline.
Her sharpest work is in the decisions that are expensive to reverse. She owns architecture review and technical AI governance: whether policy enforcement fails open or closed under load, how quickly a consent revocation propagates to an enclave holding cached state, how an enclave recovers when the instance dies mid-session, and what belongs inside the trust boundary with better isolation bought at the cost of a rebuild-and-attest cycle on every update, a release-cadence tax paid for years.
Her range is unusual in both directions. At Mondelez International she led a 41-person organization, deployed 10+ GenAI systems on private infrastructure rather than public APIs with capacity planning, serving, inference economics, model registry, a security boundary that has to hold, and shipped ~40 production models across computer vision, semantic search, digital twins, and forecasting. At Walgreens Health her models produced $17M in annualized savings analyzing and modeling the IoT Edge AI and other data. Before generative AI was a category she was building recommendation engines, a variety of ML models at Grainger, pExchange, Xen.ai, and real-time bidding on Spark streaming at 10ms latency at Centro.
Underneath is a physicist: a Ph.D. in Physics & Mathematics, grant-funded research at Argonne National Laboratory and the Carnegie Institution for Science, synchrotron spectroscopy, nanomaterials, and signal processing, two U.S. patents, and 25+ peer-reviewed publications — a background that trained her to distrust a metric until she knows where the error concentrates.
She developed and taught the NLP curriculum at Northwestern University and University of California, Berkeley, and received the 2024 GenAI Innovator Award at Mondelez.
$17M
Annualized Savings Modeled
25+
Peer-Reviewed Publications
