Vladyslav Stryzhak

• Built an LLM-powered customer-support copilot (GPT-4 + RAG on internal KB) that answers 72 % of Tier-1 tickets autonomously, cutting annual support spend $190 k. • Fine-tuned a domain-specific GPT-3.5 model (LoRA) using 40 k chat transcripts; BLEU improved +18 %. • Architected micro-service mesh (8 services, gRPC) scaling to 2.3 M req/day at p95 < 180 ms. • Introduced daily CD with blue/green & canary; mean rollback < 10 min, change-failure rate 2.1 %. • Led a 7-dev squad (0 % attrition); mentored two engineers into Staff level.

• Developed real-time fraud-detection micro-service (scikit-learn, XGBoost) that flagged 96 % of fraudulent orders with 3 % false-positive—saved $2.5 M/yr chargebacks. • Containerised training pipeline (MLflow + SageMaker) trimming model-retrain time from 4 h → 35 min. • Re-engineered auth to JWT + OAuth2; SOC 2 & ISO 27001 audits: 0 critical findings. • Conducted quarterly ML workshops; raised company-wide ML literacy score from 2.8 to 4.1 / 5.

• Built HIPAA-compliant APIs for radiology workflow platform used by 200 + hospitals, processing 3 M+ records/day. • Implemented FHIR-based interoperability services, cutting integration time with EMR vendors 60 %. • Boosted automated test coverage to 85 % (unittest+ Selenium); regression cycle shrank from 4 days to < 1 day. • On-call responder—kept mean-time-to-recovery under 20 minutes for P1 incidents.