Andrei Mosjakin

Architected and scaled AI solutions to streamline e-commerce operations, covering dynamic pricing, personalized recommendations, and inventory intelligence. Built and launched real-time voice interfaces with advanced speech recognition and synthesis, boosting customer interaction and support quality. Embedded large language models (GPT-4o, Claude, Llama) into internal platforms to power enterprise search, summarization, and intelligent automation workflows. Directed multi-agent orchestration initiatives using LangChain and vLLM, enabling adaptive query routing and automated tool-use chaining. Partnered with cross-functional teams to translate business requirements into production-ready AI features with strong user adoption. Enhanced AI pipelines for performance and efficiency through latency reduction, API request batching, and model compression strategies. Delivered and operated production-grade services via Docker, Kubernetes, and continuous integration / deployment pipelines.

Engineered and supported NLP-driven models for customer messaging platforms, specializing in intent recognition, classification, and automated response generation. Designed and optimized data pipelines to preprocess, normalize, and label large-scale text datasets for supervised and semi-supervised learning. Adapted and fine-tuned pre-trained transformer architectures for domain-specific applications, achieving higher accuracy with reduced inference latency. Implemented and deployed web APIs with Flask and FastAPI to seamlessly deliver ML services to frontend applications and third-party integrations. Collaborated closely with data engineering and product teams to synchronize ML outputs with business objectives, KPIs, and customer experience goals. Automated and monitored continuous retraining, evaluation, and deployment pipelines, enabling systems to evolve from live user feedback. Introduced robust monitoring and logging for model performance in production, ensuring reliability and early detection of drift or anomalies. Explored and validated emerging NLP techniques such as prompt-tuning and retrieval-augmented generation (RAG) to future-proof conversational AI systems.

Supported migration of monolithic legacy systems into modular, Python-based microservices, improving scalability and maintainability. Designed and delivered backend services powering product search, classification, and recommendation by combining rule-based logic with machine learning approaches. Created and automated internal utilities and scripts for collecting, transforming, and analyzing high-volume classifieds datasets. Partnered with the data science team to prototype and validate AI features such as image similarity matching and intelligent tagging systems. Enhanced API efficiency and stability through systematic refactoring, expanded test coverage, and advanced monitoring instrumentation. Implemented robust logging and alerting mechanisms to proactively detect issues and ensure service continuity.