Pavel Mikheev
Senior Backend Engineer @SoundHound AI
Signup · Get unlimited contacts
WORK HISTORY
Senior Backend Engineer @SoundHound AI
Berlin, DE
Built an LLM-based evaluation and benchmarking platform for AI agents, supporting validation of tool-calling workflows and multi-step agent execution. Implemented validation frameworks to verify agent behavior, including function-call correctness, parameter validation, and execution trace analysis, enabling controlled execution and reproducible benchmarking scenarios. Developed LLM-as-a-judge evaluation mechanisms for custom test assertions (e.g, response quality, language constraints, semantic correctness). Improved system performance by optimizing PostgreSQL schemas and queries, reducing response latency from 20s to 0.5s (40x improvement). Enhanced benchmark execution pipeline, reducing average test time from 10 minutes to 1 minute. Deployed and maintained microservices in Kubernetes, ensuring scalability and reliability of the system.
EDUCATION
Lomonosov Moscow State University (MSU)
Master's degree
Yandex School of Data Analysis
Machine Learning Course Part 1 and 2
Yandex School of Data Analysis
Linear Algebra Course
Lomonosov Moscow State University (MSU)
Bachelor's degree
ABOUT PAVEL MIKHEEV
Senior Backend Engineer specializing in high-load distributed architectures, LLM platforms and AI systems.10+ years of experience building scalable backend and production-grade AI systems, including agent-based workflows, LLM evaluation pipelines, and cloud-native infrastructure.What I focus on:• AI Systems & LLM Platforms: LLM-based evaluation and benchmarking systems for AI agents (tool-calling, multi-step workflows, LLM-as-a-judge)• High-Load Backend: Microservices and data pipelines handling high-throughput workloads (up to 10K docs/sec, 40x performance improvements)• Production Engineering: MLOps, inference services, and scalable cloud infrastructureTech stack:• Python, Go, C++, Kubernetes, AWS, Terraform, Kafka • PostgreSQL, Redis, ClickHouse • LLM systems, RAG, AI agentsExperience across SoundHound, Toloka, and Yandex. Based in Dresden, Germany (permanent residency). Working in international teams and improving my German. Remote work is supported by a well-equipped home office — with a few calm furry companions around.Let’s build great systems together!
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.