Alena Lapteva

• Built a hybrid recommendation and semantic search system using LLM embeddings, RAG architecture,
and Pinecone vector database to improve relevance of job matching and vacancy retrieval.
• Developed autonomous AI agents for job applications, automating up to 80% of the workflow.
• Designed and deployed an LLM-powered CV optimization pipeline.
• Built scalable LLM-powered services and APIs for AI-driven candidate-job matching workflows.
• Owned the full ML lifecycle from problem framing and experimentation to deployment, monitoring, and
evaluation.
• Collaborated cross-functionally with product and engineering teams to translate business objectives into
ML solutions.

• Built predictive and probabilistic models for real-time commercial decision systems, improving contract
accuracy by 12%.
• Developed large-scale ranking and optimization pipelines using behavioral and transactional data signals.
• Developed an LLM-based document parsing pipeline using OCR, NLP, and embeddings for automated
content extraction, accelerating processing speed by 8×.
• Designed experimentation and evaluation frameworks for KPI-driven decision optimization under
production constraints.
• Reduced infrastructure and data processing costs by 26% through pipeline and serving optimization.

• Improved demand forecasting accuracy by 34% across 50K+ retail locations using time-series analysis,
statistical modeling, and feature engineering on large-scale datasets.
• Built distributed ML pipelines in Spark and Databricks, reducing training time from 16h to 2h.
• Productionized forecasting models into inventory allocation systems, improving stock efficiency during peak
periods.
• Worked directly with stakeholders to translate ML outputs into operational business decisions.
• Developed scalable data processing and experimentation pipelines in production environments.

• Built high-availability distributed systems for financial institutions like Societe Générale Group
and UniCredit in regulated environments.
• Developed integration pipelines using Java and IBM Integration Bus for MQ-based systems.
• Ensured reliability, scalability, and observability for production banking systems.