04/19/25 - 04/28/25
	- Designed and printed the project poster for our upcoming demo.
	- Help with anything else the team needed, still significantly used our toolchain to compile and test code.

	- Finalized the demo setup and verified end-to-end operation on hardware.

4/12/25 -4/18/25
	- Discovered that loading our neural-network model on the FPGA’s limited RAM/BRAM wasn’t going to work.

	- Pivoted to an interactive game prototype instead.

	- I helped write the game’s RISC-V software—implementation on the DE1-SoC.

5/4/25 - 4/10/25

TFLite Micro Model
    - Converted the Micro model into a .tflite flatbuffer and then C byte array using ONNX2 library.
	- First converted the .tflite model into a onnx file and then used onnx2 to convert into C(but has floating
	point issues)

----------------------------------------------------------------------------------------------------------

3/19/25 - 4/4/25

PicoRV32 Vector extension:
    - To support the VPU, explored and ran tests on the zve32x extension of the rv32 instruction set
    - Built and used libraries using Clang/LLVM/LLD
    - Compiled sample C code and tested in simulation


03/14 - 03/18

	RISC-V Toolchain Setup:
	- Set up the RISC-V GNU Toolchain on my local machine.
	- Successfully cross-compiled a simple "Hello, RISC-V!" program and ran it on QEMU to verify the toolchain.
	
	Testing the Vosk model:
	- Successfully implemented real-time speech recognition using the Vosk model on native Windows.
	- Used Python libraries like sounddevice and vosk to capture audio from the microphone and transcribe it in real-time (testing the Vosk model, which works but not highly accurate).
	- Tested the implementation with multiple audio inputs and verified its accuracy.
	- Identified that WSL2 (Windows Subsystem for Linux) does not support direct access to audio devices, which prevented real-time speech recognition in the Ubuntu environment.
	- Explored workarounds, including using Windows for audio capture and passing data to WSL2, but decided to focus on native RISC-V deployment for the next steps.
	
	Next steps:
	- Finalize the C/C++ implementation of the Vosk model for real-time speech recognition.
	- Cross-compile the Vosk API and Kaldi library for RISC-V, or possibly explore other options depending on team's feedback.
	- I want to be able to run the Vosk model using the riscv32-unknown-elf-gcc (toolchain) compiler in real-time.
	
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Till 03/13/25

	- Worked on Proposal, Architecture review and microarchitecture review presentations.
	- Splitting work among team members.