Pinned Loading

1. flint flint Public

   Describe any workflow in plain English. Flint runs it reliably. NL→DAG engine powered by Claude, parallel asyncio, corruption detection, smart retry

   Python

2. Orbis-ML-Classifier Orbis-ML-Classifier Public

   Multilingual corporate entity classifier - XLM-RoBERTa fine-tuned on 690K samples, 98.75% precision across 120+ countries. Async batch inference at 1.4M+ entities. Associated with NYU Stern School …

   Python

3. Audit-AI Audit-AI Public

   Production-ready clinical note auditing application

   Python

4. Falcon Falcon Public

   Production ML inference platform. Multi-worker · Nginx load balancing · idempotency · exponential backoff · Prometheus metrics. Reduced p95 latency 30%.

   Python

5. StockStream StockStream Public

   Distributed real-time analytics platform. Kafka + Spark streaming at 5K+ events/sec · hybrid PostgreSQL + InfluxDB storage · Grafana monitoring · fault-tolerant consumer groups.

   Python

6. vit-inference-optimization vit-inference-optimization Public

   ViT-L/16 inference optimization - 4-bit NF4 quantization, FlashAttention-2 vs SDPA benchmarking, 40.5% latency reduction, 4x size reduction. CUDA kernel profiling with PyTorch Profiler.

   Jupyter Notebook