Looking for a job

Valentyn Yemets

@valentynyemets

Backend engineer at Stock42 who improved LLM answer accuracy from 72% to 91% with a LangGraph agent and RAG evaluation pipeline.

Argentina

Message

At Stock42, I build Go REST and gRPC microservices and connect services through Kafka. I also built a LangGraph agent over the company knowledge base to answer customer questions about products, stock, and pricing.

I improved the agent’s answer accuracy from 72% to 91% by reworking retrieval chunking and adding reranking, and reduced its hallucination rate from 14% to 3% with grounded answers and source citations. I integrated several LLM providers and built agent harnesses with tool calling, memory, guardrails, and evaluation.

Previously, at Vision Group, I built Django REST Framework APIs and asynchronous FastAPI services, and implemented background jobs with Celery and Redis. I also built a Gemma-based chat moderation classifier for a dating app.

Experience

Work history, roles, and key accomplishments

Current

Python/Go Developer

Current

Stock42

Dec 2024 - Present (1 year 10 months)

Built REST and gRPC microservices in Go following Clean Architecture and SOLID. Implemented order, user, notification, and payment modules with goroutines and worker pools, and connected services via Kafka. Developed a LangGraph agent for customer support and RAG retrieval with OpenAI embeddings and vector search, improving answer accuracy and reducing hallucination rate.

Go Gin Echo Protobuf Kafka Clean Architecture SOLID LangGraph RAG OpenAI OpenSearch Pinecone Pgvector GCP BigQuery Snowflake Streamlit Sentry

Python Developer

Vision Group

Feb 2022 - Dec 2024 (2 years 10 months)

Built REST APIs with Django REST Framework and async microservices with FastAPI and asyncio. Implemented background jobs with Celery and Redis, optimized PostgreSQL/MySQL queries, and shipped services in Docker via GitLab CI/CD. Developed an LLM chat moderation system for a dating app using a Gemma-based classifier.