Polina Reviakina

At Yolk, I build backend services for an AI sales-coaching SaaS, including transcript analysis, LLM scoring, Stripe billing, and fault-tolerant processing for around 1,000 large transcripts per day. I reduced onboarding from five steps to two with a manual upload and ingestion pipeline, and standardized structured LLM outputs to reduce parsing-related failures by 80%.

Previously at Windsor.ai, I developed integrations for 300+ marketing and CRM data sources and helped launch an LLM-powered support automation system that cut average response time by 90%. I also built a RAG pipeline over historical support tickets and fine-tuned an OpenAI model using 20,000+ curated CRM tickets.

At Timestripe, I delivered bidirectional Google Calendar synchronization for 200K users and reduced event update latency to under five seconds. I also optimized a PostgreSQL migration for 3M+ rows by 90% and expanded asynchronous processing across 17 backend workflows.

My work spans Python, FastAPI, Django, PostgreSQL, messaging systems, cloud infrastructure, and production data pipelines. I focus on clean architecture, reliable integrations, maintainable systems, and backend products that scale.