Open to opportunities

Paweł Van

@pawevan

Message

Senior AI Engineer specializing in production RAG systems—scaling latency/cost, safety, and measurable customer impact.

Poland

Message

What I'm looking for

I’m looking to own end-to-end AI/RAG systems in production—focusing on retrieval quality, observability, latency and cost optimization, safety guardrails, and measurable impact through evaluation and A/B testing, while mentoring teams and shipping reliably with CI/CD.

I’m a Senior AI Engineer with more than 10 years building and operating production ML and backend services. I “take fuzzy problems 0→1 and scale them 1→n” by owning problem framing, data readiness, document processing, embeddings, indexing, retrieval orchestration, evaluation, and production APIs.

At Diffco, I spearheaded an end-to-end RAG program—ingesting 1.2M documents, improving top-k retrieval relevance by 38%, and cutting average retrieval latency by 40% while reducing storage cost by 25%. I architected a scalable FastAPI inference/API layer on Docker + Kubernetes (EKS) to serve 500 RPS with 99.9% uptime, and I drove faster releases with automated CI/CD, plus observability with Prometheus/Grafana/Jaeger to lower P95 latency by 45% and align teams to SLOs. Earlier roles also strengthened my end-to-end engineering foundation: production RAG with Elasticsearch/Redis and A/B experimentation at Yeti, large-scale ETL and feature stores processing 5TB/month at PowerGate, and predictive + NLP modeling at Softwire.

Experience

Work history, roles, and key accomplishments

Senior AI Engineer

Diffco

Feb 2023 - Dec 2025 (2 years 10 months)

Led an end-to-end RAG program ingesting 1.2M documents and improving top-k retrieval relevance by 38%. Reduced retrieval latency by 40% and storage cost by 25%, served 500 RPS with 99.9% uptime, and cut P95 latency by 45% through AWS, FastAPI, Kubernetes, and observability/SLO instrumentation.

Lang Chain Python Retrieval Augmented Generation (RAG)Pinecone Faiss AWS S3 Lambda Glue fastAPI Kubernetes

AI/ML Engineer

Yeti

Sep 2020 - Nov 2022 (2 years 2 months)

Orchestrated production RAG prototypes for customer-facing search, integrating BERT/sentence-transformers with Elasticsearch and Redis and improving top-5 accuracy by 27%. Delivered a 15% net uplift via evaluation frameworks and A/B testing, supported 200k queries/day with median latency under 150ms, and cut embedding compute cost by 30% using batch AWS Lambda jobs on ECS.

RAG Sentence Transformers Elasticsearch Redis PyTorch Hybrid Search (BM25 + FAISS)Flask PostgreSQL AWS ECS A B Testing

AI/ML Engineer

PowerGate Software

Oct 2017 - Jun 2020 (2 years 8 months)

Built ETL pipelines with Apache Airflow, Spark, and Python to process 5TB monthly and improved model training throughput by 40% while reducing data duplication by 60% using feature stores/schemas. Migrated legacy search to Elasticsearch, increasing query throughput 3x and lowering response time by 60%, and validated data quality/lineage with automated tests achieving 99% pipeline reliability.

Apache Airflow Apache Spark ETL Feature Stores (PostgreSQL MongoDB)Elasticsearch Faiss fastAPI Data Lineage

Data Scientist

Softwire

Nov 2016 - Aug 2017 (9 months)

Built predictive models in Python and scikit-learn, increasing forecast accuracy by 21% for key metrics. Developed NLP document classification with spaCy/NLTK (120k documents, 90% precision), improved churn by 8% over 6 months via analytics, and productionized models using Docker with baseline monitoring and reproducible experiment logging.

Python scikit learn SpaCy NLTK predictive modeling NLP Docker Git

Data Analyst

Rikkeisoft

Sep 2015 - Mar 2016 (6 months)

Extracted, cleaned, and transformed transactional data with SQL and Python, accelerating monthly reporting from 5 days to 1 day and improving query performance by 60% using optimized PostgreSQL schemas/indexes. Streamlined ETL with parameterized SQL and automated refresh jobs, and trained teams on data best practices, reducing analyst onboarding time by 30%.

SQL Python PostgreSQL Tableau ETL Performance Optimization Parameterized Queries A B Test Measurement