Looking for a job

Alexander Pivovarov

@thesamedesu

Message

I build high-load backends and production LLM platforms for enterprise teams.

Georgia

Message

What I'm looking for

I'm seeking remote-first opportunities with European teams, with flexibility to relocate within the EU and full overlap with CET/CEST business hours.

At Philip Morris International, I build production LLM and data-platform products, including an agentic support assistant that handles 5,000+ conversations monthly and resolves over 30% without human escalation. I also delivered a metadata catalog spanning 300+ source systems, used by 500 daily active users for data discovery.

Previously at Zopa Bank, I built a real-time ML Feature Store serving 200+ features to 12 models at roughly 25 ms p99, and a document-recognition service processing 3,000 documents a day at 94% accuracy. I improved asynchronous pipeline reliability to 99.5% successful processing and introduced CI/CD and production monitoring.

At AUTO1 Group, I developed inventory and payment microservices serving 600 RPS at peak, reduced inventory-service p95 latency from 400 ms to 80 ms, and built Python and ClickHouse ETL powering daily dashboards. I enjoy mentoring engineers, leading design and code reviews, and establishing practical technical standards.

Experience

Work history, roles, and key accomplishments

Current

Senior Software Engineer

Current

Philip Morris International

Nov 2023 - Present (2 years 9 months)

- Designed and shipped an agentic support assistant (LangGraph, open-source LLM, RAG over Qdrant) handling 5,000+ conversations per month and resolving over 30% end to end without human escalation - around 1,500 tickets a month removed from the support queue.
- Built the retrieval pipeline - multilingual-e5-large embeddings, hybrid dense + keyword search, CrossEncoder reranking

Python Developer

Zopa Bank

May 2021 - Nov 2023 (2 years 6 months)

- Designed an ML Feature Store (FastAPI, PostgreSQL, Redis) with real-time updates over Kafka and Airflow, serving 200+ features to 12 models at ~25 ms p99 and cutting feature time-to-production from weeks to days.
- Delivered a document-recognition service (FastAPI, Kafka, S3, NLP models) automating extraction from contracts and applications - 3,000 documents a day at 94% accuracy, removing 30+ h

Python fastAPI Django Django REST Framework PostgreSQL SQLAlchemy Redis Celery Kafka Docker Airflow S3

Backend Engineer

AUTO1 Group

Sep 2018 - May 2021 (2 years 8 months)

- Developed backend microservices (FastAPI, aiohttp) for car inventory and payments, serving 600 RPS at peak across multiple European markets.
- Optimised PostgreSQL access patterns and introduced Redis caching in the inventory service, cutting p95 latency from 400 ms to 80 ms and database load by ~40%.
- Integrated external CRMs and payment gateways over RabbitMQ and Celery with idempotency and r