Hi, my name is Shaunak Sarlashkar
A Computer Engineering Student at Purdue University with interests in AI/ML, Embedded Systems, Computer Architecture, and Digital Design.
Email: me[at]shaunak[dot]dev
A Computer Engineering Student at Purdue University with interests in AI/ML, Embedded Systems, Computer Architecture, and Digital Design.
Email: me[at]shaunak[dot]dev
• Enhanced firmware image-processing path by improving reliability within a Yocto-based Embedded Linux Stack.
• Benchmarked embedded platforms with lmbench to characterize CPU, memory, and I/O behavior.
• Collaborated with firmware & image halftoning teams to validate behavior on target hardware.
• Debugged system-level issues in the lmbench suite to accurately characterize kernel-space interactions and performance metrics.
• Support graduate & undergraduate level ML and optimization assignments for ECE 570 & ECE 473 (Artificial Intelligence).
• Assist students on debugging PyTorch implementations and understand reinforcement learning pipelines.
• Grade exams and programming projects to ensure correctness classwide.
Developed an LLM-powered CLI tool to automate the pre-review process for Linux kernel patches. Provides semantic analysis on locking rationale and subsystem-specific conventions (e.g., Memory Management) before LKML submission. Complements checkpatch.pl by identifying high-level architectural issues.
Source: github.com/shaunaks-tech/kernel-lint
Engineered a 5-stage pipelined RV32I processor with full hazard detection and forwarding logic. Implemented a 2-way set associative data cache and a dedicated instruction cache to minimize memory-access stalls and improve IPC. Verified architectural performance and correctness through comprehensive SystemVerilog testbenches.
Designed an AI Accelerator featuring an 8x8 Systolic Array, AHB interface, and SRAM buffer. Developed a controller FSM to manage data hazards, reducing inference latency to under 75 cycles. Validated architectural correctness through Questasim waveform analysis and testbench simulations.
Engineered a hazard detection system on STM32F0, integrating ultrasonic sensors with UART telemetry for real-time distance tracking. Developed a non-blocking firmware architecture using ISRs for sensor polling and SPI with Circular DMA to drive a live OLED dashboard.
A React-based website that teaches Vim interactively through a LeetCode-style interface. Features a terminal emulator, lesson progression, and glassmorphism UI.
Visit at: vimdojo.app
Source: github.com/shaunaks-tech/learnvim