I help startups build firmware for their IoT products. The startup ventures I have worked with: ConcR (smart concrete sensor), Kelvyn (smart refrigerator sensors), Vitreo (smart IBC sensors), Sensry (RISC-V based universal sensor platform).• Built complete IoT solutions to collect data from various sensors and communicate with cloud services via NBIoT/LTE-M using a Nordic NRF91 chip running Zephyr RTOS. • Wrote, integrated and optimized the drivers for various modules like DS18B20/SI7055/HDC1080 temperature sensors, BME860/ZMOD4410 environment sensors, A111 radar sensor, TSL2591 ALS sensor, LIS3DH accelerometer, W25Q64JV SPI NOR flash.• Added support for various sensor data encoding formats like JSON, CBOR, Protobuf.• Set up communication with AWS IoT Core to send and receive messages over MQTT.• Automated the provisioning of fresh sensors in the factory complete with manufacturing tests, creating and flashing certificates, creating digital twins and flashing the release firmware.• Added support for firmware over the air (OTA) updates using MCUBoot. • Set up CI/CD using a Docker container to build the firmware on GitLab and also set up an • automatic release mechanism (with semantic versioning).• Involved in the development and board bring up of a RISC-V based sensor starter kit.

PROJECT: CHRISI was an integral part in the successful launch of Chris, an automotive voice assistant on an STM32 platform running a proprietary RTOS, from the prototyping phase, to manufacturing the hardware and shipping it to customers. • Successfully implemented and maintained the GUI module for Chris using TouchGFX. • Responsible for the audio module implementation on Chris including microphone capture with beam-forming and echo cancellation libraries from STMicroelectronics and a customized wake up word detector from Sensory.• Responsible for evaluating, choosing and integrating the gesture sensor used in Chris. Incorporated this research into my Master thesis and successfully defended it. • Worked on the Cypress Bluetooth module of Chris and implemented the BLE, HFP and A2DP Bluetooth profiles for communicating with the Android and iOS apps.• Fixed numerous difficult bugs in the firmware, working with the Android, iOS, and the UX teams, which led to improved customer satisfaction with Chris.• Actively participated in software architecture design discussions, sprint planning, code reviews, and knowledge sharing sessions.PROJECT: AUTOMOTIVE VOICE ASSISTANTI was actively involved in planning and designing the architecture of a Proof of Concept to port the functionalities of Chris onto a Linux PC. This PoC is going to be displayed at CES 2020.• Replicated the audio module of Chris on the PoC in 3 weeks. This involved integrating a wake up word detector library, writing an ALSA sound card driver, audio data encoding using the Opus codec and transferring the audio data over Bluetooth Low Energy to the Android app.• Coordinated with an external company, Silii, to define and implement the GUI of this PoC using QT. We were able to successfully extend the GUI of Chris and adapt it to the new requirements

• Played a key role as a part of the networking team working with the Infotainment Division. Worked on a project to bring the first Android-based Infotainment system for cars of General Motors.• Implemented of the IEEE 802.1 BA - Audio Video Bridging Protocol (AVB) for audio transport between the different components of the GM-Info3 infotainment system.• Successfully Implemented bandwidth & stream reservation in the 3PSW switch of the Texas Instruments processor• Developed a QNX resource manager for communicating with Broadcom’s BCM89501 switch for diagnostic and software download support.• I was awarded a US patent (US20160191597) along with my manager for our idea about developing an AVB diagnostics protocol

• I successfully completed my Bachelor thesis on the study of the efficiency of various supervised classification algorithms for land surface features and their application on hyper-spectral images. • Carried out a performance evaluation to find out the effect of dimensionality reduction on the classification efficiency.

• Worked on the firmware for a robot being designed at the institute. The robot was mounted with a leanXcam and would drive down a corridor keeping a constant distance from the wall.