Bruno Wassermann

CV

A condensed version of twenty-three years. For the full document, email me.

IBM Israel Research Lab Research Staff Member · 2013 – present

  • 2025 –Agent cost optimisation — technical team lead. Uncertainty-aware dynamic routing to improve the cost–quality trade-off of LLM agent flows; controlled experiments to establish the return; generalising the approach into llm-d.vLLM · llm-d · Go · Python · OpenShift · LiteLLM
  • 2024 – 25LakeGuard — technical team lead. Access control for structured and vector data in Gen-AI lakehouses: one model for heterogeneous ACLs, fine-grained enforcement across multiple query and storage systems, benchmarking what each filtering strategy costs in latency.Presto · Iceberg · Milvus · OpenSearch · Postgres · Spark
  • 2024ARISE — models predicting the resource use, latency and throughput of LLM fine-tuning and inference jobs, so GPU type, count and batch size can be chosen rather than guessed.scikit-learn · XGBoost · pandas
  • 2023 – 24Hybrid data flows — technical team lead. Optimising data processing that spans local clusters and cloud.Kubeflow Pipelines · Apache Calcite · Substrait
  • 2021 – 24Hybrid cloud data lifecycle management — technical lead. Policy-driven expiry, archival and transfer of data across Kubernetes and OpenShift clusters, from factory floor to cloud.Quarkus · Java · Kubernetes · MongoDB · Spark
  • 2016 – 22AIOps for IBM IaaS — tech lead. Anomaly detection and root-cause analysis over logs and metrics for site and network reliability engineers; adopted in production by the team running the IBM Cloud network in 2020.Python · Kotlin · Vue.js · Elasticsearch · Kafka · Grafana
  • 2018 – 19Watson Machine Learning Platform — tech lead. Automated release-candidate verification for the NLU training pipeline behind Watson Assistant, so a continuous deployment cannot quietly degrade a customer's chatbot.Python · applied statistics
  • 2014 – 17Watson Developer Cloud — tech lead. Core distributed-systems infrastructure from the first open beta through significant growth; SRE for core services; lifecycle automation for tens of ZooKeeper clusters; close to zero downtime with a team of two to four.Bash · Python · Ansible · ZooKeeper · Java · ELK · Grafana
  • 2013 – 14Cloud enablement and mobile backend-as-a-service — making a single-tenant mobile application platform run as a cloud-native, multi-tenant service.Java · JPA · JAX-RS · WebSphere Liberty

Earlier

  • 2011 – 13Duco (then Model Two Zero), London — contractor. Verification rules and tests for SWIFT message types.
  • 2008IBM Research, T. J. Watson, New York — intern. Structured change management for cross-domain service mashups. Two conference papers; Best Student Paper at ICWS 2009.
  • 2006 – 08Eclipse Foundation — committer on the Eclipse BPEL Designer. Built the runtime integration framework that let third-party BPEL engines plug into the designer.
  • 2004 – 06University College London, Software Systems Engineering — research assistant on OMII-BPEL. Built Sedna, a graphical environment for computational scientists to express experiments as Grid workflows; used in production for several large-scale workflows.
  • 2003 – 04Goldman Sachs International, London — programmer analyst, asset management applications. Preceded by a summer internship in 2002.

Education and recognition

  • 2012PhD, Computer Science — University College London. Data-Driven Detection and Diagnosis of System-Level Failures in Middleware-Based Service Compositions.
  • 2007Understanding Entrepreneurial Management — London Business School.
  • 2003MSci, Computer Science with Cognitive Science — University College London.
  • —Six granted US patents. Best Student Paper, ICWS 2009. Recognised inside IBM for the revenue contribution of the Watson Developer Cloud work, the business impact of the cloud-enablement work, and contributions to Conversational Technology.
  • —German (native), English (fluent), Hebrew (conversational).