Liudmila Lapitskaia

I’m an Applied ML Engineer who turns ML models into things that actually work in production—built for reliability, observability, and measurable user impact. With 6+ years of experience scaling ML systems, I’ve delivered solutions across recommendation, search relevance, and document intelligence for large audiences.

At Delvify (Dec 2024–Present), I built a document intelligence product from scratch for handwritten Japanese archival text forms—delivering end-to-end data/ML pipelines and a web platform to view, edit, and verify results. I also built fault-tolerant LLM OCR batch pipelines (~90% accuracy) processing 50k+ documents with observability, progress tracking, and cost optimizations.

Previously at PayPal (Sep 2019–Jun 2023), I built and maintained 100+ Spark ML pipelines processing 1TB+ of data daily, powering feature extraction, data enrichment, training dataset generation, and ranking/recommendation algorithms that improved search relevance for 100M+ users. I led MLOps/Data Engineering to scale production pipelines from 3 to 100+ and delivered tooling that cut deployment time to seconds while reducing disk and namespace quota usage by ~99%, with end-to-end observability via Splunk + SignalFX.

Earlier at Jetlore (Oct 2017–Sep 2019), I drove real-time user segmentation insights using AWS Athena and migrated pipelines from Apache Cassandra to AWS Athena, cutting cost by ~20%. I also delivered end-to-end full-stack features and led a team rewrite from vanilla JavaScript to React in 2 months with 0 bugs and near-100% test coverage—always shipping high-quality results while keeping engineering fundamentals tight.