Pinned Loading

1. Parallel-BPE-Tokenizer Parallel-BPE-Tokenizer Public

   High-performance GPT-2-style BPE tokenizer in C++ with parallel batch encoding, thread-pool execution, thread-local caching, and benchmark-driven comparison against tiktoken and GPT2TokenizerFast

   C++

2. CPUinference CPUinference Public

   End-to-end LLM inference on CPU: API serving, streaming, benchmarking, memory analysis, and model compression(quantization).

   Python

3. FlashSeatReservation FlashSeatReservation Public

   Flash-sale seat reservation backend (holds + bookings) built with Spring Boot + PostgreSQL, with concurrency-safe constraints, Flyway migrations, Docker, GitHub Actions CI/CD, and Azure Container A…

   Java

4. gpu-profiling-cuda-kernels gpu-profiling-cuda-kernels Public

   GPU profiling suite & CUDA kernels on A100 80GB — ResNet-50 benchmarks, Nsight Systems profiling, tiled matrix multiplication with shared memory

   Python

5. Hardware-Aware-Training-Time-Throughput-Prediction Hardware-Aware-Training-Time-Throughput-Prediction Public

   Hardware-aware CNN training performance predictor for CIFAR-10 on NVIDIA A100—learns sec/epoch from config features and derives throughput (images/sec) from predicted time.

   Python

6. Doc-Rag-Agent Doc-Rag-Agent Public

   Doc RAG Agent is a custom lightweight document Q&A system built from scratch that retrieves relevant chunks from PDFs, answers using only that evidence, and enforces verified citations (or abstains…

   Python