realtime-inference

FlashRT is a high-performance realtime inference engine for small-batch, latency-sensitive AI workloads. The flagship integration is production VLA control for Pi0, Pi0.5, GROOT N1.6, and Pi0-FAST. Also support llm e.g, qwen3.6-27B

A real-time Retail Shelf Monitoring System using computer vision and machine learning. Detects out-of-stock products, misplaced items, and ensures planogram compliance through intelligent video analytics and a desktop management interface.

GPU-resident TensorRT pipeline for FoundationStereo. 2.7x faster on Jetson Orin, bit-identical FP16, <0.2% INT8 error.

Real-time Threat Classifier deployed on SageMaker via AWS CDK. Full MLOps lifecycle: secure VPC deployment, DistilBERT inference, P95 latency enforcement, and CloudWatch-driven automated rollback.

Webcam'den 30 Türk İşaret Dili kelimesini gerçek zamanlı tanıyan bir uygulama. MediaPipe ile vücut ve el iskelet noktaları çıkarılıyor, kayan pencere üzerinde LSTM + Masked Attention modeli çalışıyor, majority voting ile kararlı tahmin üretiliyor.

Real-time stress detection wearable using ESP32 + multi-sensor HRV (ECG / PPG / GSR / temp / IMU), SVM trained on WESAD, FastAPI inference service, and React dashboard. LOSO accuracy 0.891 across 16 subjects.

Low-latency microservice using Causal Inference (Calibrated LightGBM T-Learner) and Risk-Adjusted Expected Monetary Value (EMV) decision gating to optimize e-commerce gross margins. Features dual Client and Server-Side tracking, zero-Pandas NumPy inference, async Pub/Sub streaming to BigQuery, FastAPI, and Cloud Run CI/CD.

Full Tracking System create to apply advanced software engineering topics with ofcourse other roles