Amir Kurmangaliev

Senior Backend Developer

ACT

Financial AI Assistant and Construction Monitoring Platform for Linear Objects, oil & gas

Built a financial-document RAG assistant and modernized a construction-monitoring platform; shipped new services and cut CI build time by 35%.

• Built a Python/FastAPI backend from scratch for a conversational assistant analyzing financial and construction documents (PDF, Excel, KS-2, GRAND-Smeta): document upload, parsing, chat sessions, and real-time streaming responses. • Designed the full RAG pipeline: document parsing via Docling, structure-aware semantic chunking, metadata enrichment, and vector embeddings in Qdrant; implemented semantic search with metadata filtering scoped per user.

Python Developer

Sberbank

Salary card issuance platform

Improved reliability and security of a high-load card platform, cutting frontend round trips by 40% and Kafka lag from minutes to seconds.

• Implemented a rate limiter on critical endpoints, eliminating downstream service overload from burst traffic. • Built an internal BFF service on Node.js to aggregate data between card-platform microservices, reducing frontend round trips by 40%. • Resolved chronic Kafka consumer rebalancing issues (aiokafka), reducing message lag from minutes to seconds. • Closed 30 security vulnerabilities identified in an internal audit through dependency updates, input-validation hardening, and removal of unsafe libraries.

Python Developer

DomClick (Sber)

Mortgage application platform

Built mortgage-platform microservices and integrations, reducing average application processing time by 25% through SQL optimization and caching.

• Built and maintained async FastAPI microservices, integrating with a credit bureau (BKI), property valuation service, and core banking system (ABS) via async httpx with retry and circuit-breaker logic. • Implemented credit-decision calculation logic and mortgage deal parameters using multi-step Pydantic v2 validation pipelines. • Reduced average application processing time by 25% through SQL optimization and Redis caching. • Implemented document reuse for refinancing cases, reducing manual manager steps per application.

Python Developer

Emoushon

EdTech platform

Owned backend delivery for a greenfield EdTech platform, including services used by 2,000+ daily users; test coverage reached 80%+ and production bugs fell 60%.

• Built a scalable file-management service for upload, versioning, and deletion, serving 2,000+ daily users. • Built a notification system for parents covering missed classes, low progress, and upcoming payment deadlines via RabbitMQ and email/SMS gateway; reduced administrative workload at education centers. • Developed an automated learning-progress module that combined attendance, completed assignments, and test results into a unified score, enabling teachers to track performance without manual data collection.

Bachelor's, Computer Science & Engineering

ASTU