Aleksi Svanidze

I've benchmarked speech models and built large-scale speech-data pipelines at Huawei's Amsterdam R&D Center, evaluating TTS, STT, ASR, and voice-cloning systems across accuracy, latency, and robustness.

I architected Vortex, a distributed search engine using Go workers, RabbitMQ, Redis, PostgreSQL/pgvector, Python embeddings, Docker, and Prometheus/Grafana. I built resilience into crawling through deduplication, rate limiting, retry backoff, robots.txt compliance, and SSRF protection.

At Scale AI, I solved and annotated Olympic-level algorithmic problems for LLM training data with a 98% acceptance rate. I was promoted to mentor 700+ contributors and then to Global QA, reviewing correctness, originality, and quality across the pipeline.

My work also includes an O(1) algorithm for generating the Hoffman-Singleton graph and HDO Tracer, a Python engine that recovers disease-propagation paths across hundreds of nodes in under 50ms. I'm a Codeforces Expert, an ACM-ICPC Northern Eurasia Finals participant, and a Google STEP offer recipient.