·
Builders & Technical Leaders
Janni Turunen
Head of Engineering |Fuzu
Technology, design, and ML leader with over a decade of experience building internet-scale products. Former CTO at Fuzu and Head of Engineering & Analytics at Custobar. Design thinking practitioner who applies ML to create added value, working across Ruby, Python, TensorFlow, Keras, and Kubernetes.
Biography
Deep Research
Janni Turunen is a Helsinki-based technology, design, and ML leader. Turunen studied Politics and Development Studies at SOAS University of London (1997-2000) before pivoting to internet and engineering roles. After leading engineering and analytics at Custobar and holding CTO/Partner roles across Solita, Aitomation, and D2 Solutions, Turunen joined Fuzu Ltd as Head of Engineering. At Fuzu, Turunen leads development of an AI-powered career platform serving 2.8M+ users across East Africa. Turunen publishes optimized open-source models on HuggingFace, including INT8-quantized Qwen3 embeddings, and builds Rust-based developer tools like vipune (semantic memory) and oo (context compression for AI coding agents).
ONNX model optimization for production deploymentINT8 quantization of transformer embedding modelsText Embeddings Inference (TEI) pipeline optimizationCPU-optimized inference for embedding modelsQwen embedding model fine-tuning and deploymentRust-based systems programming for AI infrastructureInfrastructure as Code (IaC) for AI platform scalingWhite-labeled AI recruitment platform architectureAI-powered job matching and candidate screening systemsTypeScript integration for AI API services
Timeline
10 Research10 total
2025
2024
2022
2021
1997
1992
Key Contributions
Qwen3-Embedding ONNX INT8 Quantized Models
Created INT8 quantized ONNX versions of Qwen3 embedding models optimized for Text Embeddings Inference (TEI) with CPU acceleration. These models achieve 8x size reduction (0.56GB vs 4.7GB) and 2-4x faster CPU inference while maintaining accuracy.
oo - Context-Efficient Command Runner for AI Coding Agents
Developed a Rust-based tool that intelligently compresses command output for AI coding agents, reducing context token waste by classifying and summarizing verbose build/test output while preserving essential information.
Fuzu AI Recruitment Platform Architecture
Led engineering for Fuzu's white-labeled AI-powered recruitment platform serving African markets, implementing scalable job matching, candidate screening, and career development systems that process thousands of job applications daily.
TEI Porter Deployment System
Created deployment infrastructure for Text Embeddings Inference models using ONNX runtime with AMD-optimized builds, enabling efficient CPU inference for embedding models in production environments.
Builders & Technical Leaders/Janni Turunen
All Profiles