Aim of this project is building a revolutionary Elastic Acceleration for AI. A SaaS product enabling more DL/ML training and inference, In less time, On less hardware, and across any AI chips(GPUs, FPGA, TPU, specialized chips).

* To utilize nvidia gpu's effectively for tensorflow machine learning framework, we implemented below mentioned features.

1) slicing of nvidia gpu resources(gpu memory,BlockPercent,Smipercent).

2) Auto allocation of gpu memory for tensorflow application.

3) Waiting for gpu resources.

4) Multiple gpus slicing for tensorflow application.

5) Multi instance feature(kubernetes).

* By using above mentioned features nvidia gpu's utilization got improved a lot.

Aim of this project is providing provision to DL/ML training and inference across any clouds and on less nvidia cuda gpu hardware.

The GPU Virtualization Service presented in this work tries to fill the gap between in-house hosted computing clusters, equipped with GPGPUs devices, and pay-for-use high performance virtual clusters deployed via public or private computing clouds. RVirtuS allows an instanced virtual machine to access GPGPUs in a transparent and hypervisor independent way, with an overhead slightly greater than a real machine/GPGPU setup. The performance of the components of gVirtuS is assessed through a suite of tests in different deployment scenarios, such as providing GPGPU power to cloud computing based HPC clusters and sharing remotely hosted GPGPUs among HPC nodes.

* Virtualized nvidia cuda based gpu's for high performance compute(Hpc) and cloud computing.

* Real machines can access remote GPUs and access of more clouds from remote system.

* generic virtualization and sharing system enables thin Linux based virtual machines touse hosted devices as nVIDIA GPUs.

* By virtualizing nvidia cuda based gpu's,we are able to scale AI applications.

* Application will run from cpu machine(frontend), it will contact gpu machine (backend) for gpu resources.

Aim of this project is to improve DL/ML training performance in cpu using persistent memory(pmem).

* By creating persistent memory device(/dev/pmem) from Random access memory space and implemented read and write apis with the help of pmdk(persistent memory development kit) and pmdk driver.

* Provides access latencies less than those of flash SSDs.

* Real-time access to data; allows ultrafast access to large datasets.

* Acheived better performance while reading data sets in tensorflow machine learning applications using pmem.

Aim of this project is controlling of home switch board from anywhere in the world and saving of power consumption.

* By using smart wireless switch,we are controlling switch board connected in home with help of android phone.

* smart wireless switch is iot product, it will sit inside the switch board.

* Android application will take control signal from user and it will update to cloud. From cloud

wireless switch will take control signal and it will control the switch board respectively.

* In this application mosquitto protocol used to establish communication wireless switch(pubisher) to cloud(broker) and cloud to Android application(subscriber).

* Installation of smart wireless switch is very easy.

* Smart wireless switch used in home appliances, industry automations.

Aim of this project is collecting data from multiple end devices and publish data to cloud.

The QCA4531 is a two stream (2x2) 802.11b/g/n single-band programmable Wi-Fi System-on-Chip (SoC) for the Internet of Things (IoT). This low-cost turnkey solution combines high performance connectivity capabilities with a user-programmable Linux OpenWrt environment and is designed to serve either as a feature-rich IoT node or as a hub to support an IoT ecosystem.

* Gateway has operating system QSDK(qualcomm software development kit).we made it up qca4531 board by flashing kernel,rootfs and initrd images generated from QSDK.

* qca4531 soc chip has wifi,bluetooth,zigbee modules.we implemented application by utilizing all modules(wifi,bluetooth,zigbee) and cross compiled gateway application by using QSDK compiler.

* Gateway is qca4531 and end devices with esp,arduino.

* In this, Gateway will communicate with cloud(cloud MQTT,AWS,GCP,Digital ocen or microsoft Azure) and all end devices are communicates with respective services in gateway.

* Gateway will publish data to cloud and from cloud android application will subscribe data for user visualizations.

* used for home applications,industry applications.