backend engineer · python · belgrade

I turn raw data into systems you can actually use.

Six years of indexers, SDKs and APIs for on-chain data — the same schema-design and retrieval instincts now go into LLM/RAG infrastructure. Every system snaps together from small, well-made pieces.

Things I've built

01 · sqd · python

Python SDK for SQD Portal

The client library that streams indexed on-chain data straight into your Python pipeline. Designed the API surface, shipped it, wrote the docs.

02 · dexguru · python · clickhouse

EVM ingestion pipeline

Indexes on-chain events across multiple EVM chains — prices, trades, balances, new assets. Query, schema and caching work made retrieval 50% faster.

03 · dexguru · fastapi

Meta Aggregation Trading API

Many trading providers behind one interface. Every request takes the most profitable execution route — a non-deterministic system with many moving parts.

04 · side project · pgvector

RAG over on-chain data

Ask questions in plain English, get answers from token and contract metadata via an LLM.

05 · side project · ios

FocusCall

Calls you by phone instead of sending a push notification you'll swipe away. Built for ADHD brains that ignore notifications. Live on the App Store.

06 · side project · llm

imediq

AI-assisted interpretation of medical test results. Biomarker tracking normalized across labs via LOINC.

More on GitHub github.com/shestakoven ↗

now

Right now: backend engineering on an LLM/RAG platform — a vector database product, at scale. Same instincts as always, new domain: schema design, and retrieval that's fast and trustworthy. Before that — Flink jobs, ClickHouse sinks and a Python SDK at SQD, and five years of DeFi data engineering at DexGuru. On the side: FocusCall, an app that calls you instead of nagging you with a notification, and imediq, AI-assisted reading of medical test results.

tools

Python · TypeScript · Java (Flink) · SQL · Solidity
FastAPI · GraphQL · PostgreSQL · ClickHouse · Redis · RabbitMQ · Kubernetes