Oleksii Kotliar

Open to B2B engagements · Lisbon, EU

Oleksii (Alex) Kotliar

Staff AI Infrastructure Engineer

I build AI infrastructure — retrieval, agent memory and evals — on top of distributed data systems (Python, Spark, Kafka, JVM). 13 years of experience, independent contractor, overlapping with US and EU hours.

me@oleksiikotliar.dev · linkedin.com/in/oleksii-kotliar · github.com/zfrgt

Email me CV (PDF) LinkedIn GitHub

Highlights

  • 1B/dayad events through a Scala/Spark/Python/Go data platform I led — in production for 5+ years.
  • 2.5M req/sthroughput on a high-performance Scala/Spark data platform I contributed to.
  • −50%storage cost on several of Lyft's petabyte-scale datasets, with integrity and granularity preserved.
  • −75%resource usage after rebuilding a low-latency processing service on Akka/Pekko.
  • −80% MTTRafter redesigning incident response while managing a team of 7 engineers.

Now

Contracting on AI retrieval and agent infrastructure. Focused on retrieval quality, eval pipelines, and making LLM-powered products reliable in production.

Most interested in the plumbing underneath agents: memory with decay, reranking, and the data pipelines that feed retrieval.

Experience

  1. Staff Software Engineer · Confidential client

    – Present

    Remote · full-time May–Aug 2025, B2B contractor since Aug 2025

    • Designed and built retrieval (RAG) infrastructure for production AI agents, covering ingestion, indexing, retrieval and reranking.
    • Built agent memory components, including relevance modelling that strengthens and decays over time.
    • Built evaluation pipelines for retrieval quality using synthetic golden datasets and LLM-as-a-judge.
    • Contributed to security reviews, mentored teammates and drove cross-team architectural alignment.
    • Earlier in the engagement: data pipelines and AI-driven data enrichment.

  2. Engineering Manager · Nexla

    –

    Lisbon, Portugal (remote)

    • Managed a team of 7 engineers building and maintaining a high-performance data ingestion/delivery system.
    • Established a culture of transparency and trust that measurably improved team morale and productivity.
    • Designed and rolled out a new incident-response process that cut response time by over 80%.

  3. Senior Software Engineer · Nexla

    –

    Lisbon, Portugal (remote)

    • Designed and built a highly available, fault-tolerant data processing pipeline on Apache Spark using Nexla's core platform and modern cloud infrastructure.
    • Built a low-latency data processing service on Pekko, cutting resource usage by at least 75%.

  4. Data Engineer, L6 (Senior/Staff) · Lyft

    –

    Lisbon, Portugal (remote)

    • Led a team of engineers delivering datasets used by roughly half of the company's data science teams.
    • Reduced storage cost and cut several datasets' size by 50% while preserving data integrity, relevance and granularity across the petabyte-scale storage layer.
    • Stack: Python, SQL, Apache Airflow, Boto3, AWS.

  5. Senior Software Engineer · Sigma Software

    –

    Odesa, Ukraine · remote from the EU since early 2022

    • Contributed to high-throughput data platforms in Scala/Spark (~2.5M req/s) and an AdTech ETL pipeline on Hadoop (Spark, Kafka, MemSQL).
    • Led the Clean Data Platform (Oct 2019 – Nov 2021): Scala/Python/Go on Spark, processing up to 1B ad events/day with ELK as a low-cost data lake for ad-hoc analysis.

  6. Software Engineer · Sigma Software

    –

    Odesa, Ukraine

    • Migrated a major customer to the next product version, owning migration issues end-to-end.
    • Contributed to "Gaming Platform", a framework handling game-client communication, transactions, logging, sessions, game history and portal/widget delivery, letting game studios focus on game logic (Java, OSGi/Apache Felix, Apache CXF, Spring, Hibernate, Akka; BDB, H2, Oracle).

Writing

  • Nexla Spark ETL On Databricks · Nexla blog

    article

    How Nexla's Spark ETL integration with Databricks works: sources and transformations, cluster lifecycle, and Delta table output.

Skills

AI / LLM infrastructure
RAGRerankingAgentic memoryLLM evalsVector searchEmbedding modelsImage processing & OCR
Languages
PythonSQLJavaScalaGroovy
Data & streaming
Apache SparkKafkaFlinkAirflowPresto/TrinoAkka/PekkoBeam
Storage & databases
PostgreSQLDynamoDBClickHouseVerticaMySQLBDB/H2/Oracle
Cloud & infra
AWSGCPHetznerKubernetesECS/EKSArgoCDDocker
Tooling & delivery
GitGitHub ActionsJenkinsPoetryBoto3PandasNumPyShellClaude Code
Leadership
Engineering managementMentorshipCode & design reviewProcess improvementCross-team technical leadership

Education

MSc in Computer Science, Odesa National Polytechnic University

– · Odesa, Ukraine

Thesis: OCR with image pre-processing, segmentation and stroke-width transform.

Degree recognised in Portugal (DGES, 2024).

Languages

  • English C2
  • Ukrainian Native
  • Russian Native
  • Portuguese B1