MAKSIM VIALYKH

RemoteAI-native ERP integration platform. Sole technical contributor responsible for platform architecture, AI systems design, and production delivery. Designed and built Scotty Bonnet, a multi-ERP translation layer using Python/FastAPI and AWS serverless infrastructure (Lambda, DynamoDB, SQS) enabling plug-in integration adapters for multiple ERPs. Engineered Scotty Coder using Claude API with per-ERP Jinja prompt templating to generate ERP-specific integration code. Built an ML-based UOM predictor (~97% accuracy) using scikit-learn to normalize heterogeneous inventory schemas. Architected Dumbledore, a no-code wizard builder for configuring integration workflows without engineering involvement, and implemented Carmen for operational visibility, audit logs, and connector health. Established environment separation, DynamoDB model decoupling, and Windows connector code signing to close production reliability and compliance gaps.

Re-architected on-prem Certificate Authority Adapter, improving authentication reliability from 95% to 99.9%. Increased identity server throughput by 50%, enabling millions of authentication requests daily. Introduced CI/CD and automated QA gates, reducing production issues by 70%.

Lead 20+ engineers across iOS, Android, macOS, and Windows; scaled from 8 developers; mentored future leaders. Introduced Product Factory model aligning design, QA, and product, reducing delivery bottlenecks by 40%. Established predictable monthly release cadence with telemetry-driven QA, achieving 95% on-time releases. Built unified AI-powered QA automation framework, reducing regression testing time by 60%. Delivered secure identity apps for Fortune 500 clients integrating FIDO2, PKI, and SmartCard authentication. Hands-on with architecture, SDK design, and cross-platform app delivery.