Amsterdam, North Holland, Netherlands
LLM Inference Optimisation, Multi Agent orchestration for enhancing travel experience at scale
AI agents x100
Key focus areas - > Solution Architecture - Creation, Guidance and Advice > AI Engineering > Generative AI, LLMOPS > AI Strategy vision
• LLM-evaluations for RAG based chatbot - Developed online and offline evaluation architecture achieving xx% recall across 2+ European countries - Implemented language-agnostic synthetic data generation for enhanced classifier performance • Performance Optimisation - Engineered custom asynchronous processing for LLM operations - Implemented advanced prompt engineering and context caching mechanisms - Designed cost-effective token management and batch processing solutions • MLOps & Infrastructure - Architected LLM architecture pipelines using GCP, Kubeflow and Vertex AI - Implemented robust CI/CD workflows with automated testing - Developed scalable monitoring and alerting systems using BigQuery
• Enhanced ML-classification model to detect and prevent customer fraud by refactoring ETL process and parallelising code components. • Conducted experiments to optimize computational challenges for managing data of 100+ million transactions. • Standardized code for deployment and experimented with MLflow for improved efficiency.
- Developed ML- classification models in the domain of customer interaction such as churn prediction, propensity to buy and personalisation models for the entire 3 million retail customer base in Google Cloud Platform (GCP). - Developing automated MLOPS pipeline utilising vertex AI GCP platform. Setup personal initiatives towards optimising the current pipeline by creating a scheduler, allowing automated pipeline runs every month. Stack : Google Cloud Platform · sklearn · CI/CD.