Andrei Pustovoi

Co-developed a study-processing pipeline for a healthcare-AI teleradiology platform, executing 3.7M+ AI-inference runs across 33 models on 500K+ studies. Built a reporting platform from MVP to production in 4 months and designed a self-serve configuration service that cut client onboarding from weeks to days.

•    Co-developed the platform's study-processing pipeline, a Celery/RabbitMQ conveyor that has executed 3.7M+ AI-inference runs across 33 production models on 500K+ CT/MRI studies, sustaining 950+ studies/day behind a gateway serving 14M HTTP requests/month.
•    Co-built a teleradiology reporting platform from MVP to production in 4 months: FastAPI backend, SuperTokens auth, and an 8-stage workf

•  Built a multi-provider money-transfer comparison engine: parallel rate polling (asyncio.gather) across 4 external FX APIs with fallback chains and multi-layer caching, ranked by a weighted algorithm across rate, fees, speed, and reliability.
•  Reverse-engineered the undocumented Binance/Bybit P2P APIs and sharply cut upstream request volume: strict Pydantic validation against silent API drift,

•  Led a team of up to 5 engineers building an ad-management platform for Wildberries and Ozon, Amazon-scale e-commerce marketplaces: directed architecture and delivery, cutting option-search time from weeks to minutes (a 99% reduction).
•  Owned hiring and people management: sourcing, technical interviews, offers, performance reviews, regular 1:1s, and onboarding.
•  Designed and built an LLM Aut

•  Built the document-generation engine of a low-code B2B platform: financial documents and scientific reports assembled by up to ~50 client-embedded scripts per document, with numeric and formatting correctness as a hard requirement.
•  Made 100-document batch export 9× faster (~3 h - ~20 min): parallel per-document Celery tasks, one hand-written SQL query per batch instead of an ORM N+1 cascade,

•  Led a 4-person team (3 developers + 1 designer) delivering the college’s two core systems: attendance tracking and the public scheduling site (Django, Plotly, PostgreSQL).
•  Proposed and drove adoption of role-based attendance tracking - own initiative pitched to the administration - eliminating a 20+ hour/week paper routine with live Plotly analytics.
•  Scaled the scheduling site to 2,000 co

•    Built a real-time Discord voice AI assistant: a voice capture → Google Speech-to-Text → OpenAI GPT → TTS playback pipeline with a custom audio-frame sink.
•    Shipped a Telegram CPA/affiliate bot: a full campaign-management backend and cryptocurrency-payment subscriptions (NOWPayments).
•    Delivered end-to-end for a range of clients: an Epic Games storefront notifier bot, a Django/Ajax e-c

Pursued higher education in Computer Science, focusing on software development and systems design.