Александр Шульгин

I’m a backend/LLM engineer focused on LLM in production, building multi-LLM pipelines that run 24/7 for commercial and client systems. I handle the full loop solo—from problem statement to production—covering scoring, classification, content generation, OCR translation, and data validation.

My systems are built for reality: latency-critical paths go to Rust, while Python handles orchestration and reliability. I’ve achieved 99%+ uptime and reduced latency from 290–596 ms to 18–41 ms, with cross-provider retries/fallbacks and guards that prevent templated replies.

In my recent independent work, I shipped Telegram and automation products (e.g., Crime Token Scan, HH AutoApply, Manga Translator) and production Web3 workflows (e.g., Transaction Guardian for real-time EVM monitoring, plus Polymarket trading infra with a Rust critical path). I also translate R&D into results—like a ×13 Rust hot-path speed-up and +59% PnL versus manual tuning—because I optimize by measurement, benchmarks, and failure-proofing.