Open-source LLM routing evaluation platform benchmarking cost, latency, quality, and safety across routing strategies, with a focus on Turkish banking.
-
Updated
Nov 11, 2026 - Python
Open-source LLM routing evaluation platform benchmarking cost, latency, quality, and safety across routing strategies, with a focus on Turkish banking.
RAG evaluation framework. Measures retrieval precision/recall/F1, checks answer faithfulness via embeddings and LLM judge, profiles latency, and compares configs. Runs in CI.
Fault-tolerant distributed stock trading system in Python — Paxos consensus (Proposer/Acceptor/Learner, Prepare/Promise/Accept/Learn) across 3 replicated order service nodes for consistent transaction ordering under concurrent writes, partial failures, and leader crashes; LRU cache, Docker Compose, AWS load testing
CPU-only conversational voice assistant with VAD-based speech segmentation, streaming LLM response generation overlapped with concurrent TTS synthesis via producer-consumer queues. Batch STT (Whisper). Fully instrumented with per-stage latency benchmarking.
Quantization profiling, inference latency benchmarking, and Random Forest execution time prediction for Google Gemma on constrained hardware.
When does a learned model stop beating physics? Measuring the noise threshold where an 18K-parameter trajectory model loses to a Kalman baseline — and what it costs to serve at 10 Hz.
A Jev-inspired System One AI model built from scratch for fast, calibrated probabilistic decisions which includes parallel inference, agent security classification, and accuracy/latency benchmarks against pretrained transformer LLMs.
To associate your repository with the latency-benchmarking topic, visit your repo's landing page and select "manage topics."