Computer Engineer

Yav Rohatgi Software Engineer.

Software Engineer at Sonet.io and a recent Computer Engineering graduate from UMass Amherst, with a strong interest in |

Yav Rohatgi
Experience

Where I've worked

Software Engineer

June 2025 – Current
Python
PowerShell
Cloud Infrastructure
  • Developed an agentic AI solution using browser automation to read and drive web based application screens, handling logins, data entry, and workflow execution, achieving 80% accuracy on PMS workflows
  • Automated VM image preparation using CI/CD pipelines, including WinRM validation, sysprep, VM generalization, and Shared Image Gallery (SIG) publishing, streamlining and speeding up image release cycles
  • Built custom browser automation actions using Chrome DevTools Protocol for precise coordinate based clicking and typing on canvas based UIs, enabling reliable automation on web applications, where DOMs fail

Undergraduate Teaching Assistant

Sept 2023 – Dec 2024
MATLAB
Excel
Python
  • Helped 150+ students by conducting weekly office hours and debugging their Python & MATLAB assignments
  • Assisted students by breaking down assignment problems into clear steps and explaining the question
  • Handled course logistics such as grading, student coordination, and communicating common issues to instructor

Software Engineer Intern

May 2024 – Aug 2024
PowerShell
Bash
Kubernetes
  • Cut manual setup time by 7% by engineering PowerShell scripts to automate installs on 50+ remote machines
  • Increased log access speed by 8% by building Kubernetes monitoring scripts for 10+ applications across 3+ clusters
  • Improved system and storage efficiency by 5% through logout scripts that deleted guest-created data

Software Engineering Intern

May 2023 – Aug 2023
Natural Language Processing (NLP)
Python
SQL
  • Boosted sentiment classification accuracy by 25% by developing a Python program for 50K+ product reviews
  • Customized VADER to identify key product features tied to profit and market shifts across 100+ smartphones
  • Delivered insights to product teams that influenced market strategy and product improvement decisions
Projects

Things I've built

Sign Language Detection Gloves

Smart gloves built to recognize sign language gestures using machine learning and IMU sensors, developed under the guidance of Professor Hossein Pishro-Nik.

Machine Learning
Python
C++
View details Code

LLM-Based Math Tutoring System

Research project under Professor Lixin Gao, focused on enhancing mathematical reasoning capabilities in large language models.

LangChain
Prompt Engineering
Hugging Face
View details

Compact LLM Architecture

Efficient implementation of language prediction models using JAX for optimized computation.

Machine Learning
JAX
View details

Thread Library & Disk Scheduler

Low-level implementation of threading and disk scheduling systems in C++.

Operating Systems
C++
Multi-Threading
View details

Token Manager and Holder Contracts

Ethereum smart contracts for secure token management with ERC223 compliance.

Solidity
Ethereum Smart Contracts
View details

Mixed Reality Research

Implemented shared object interaction in mixed reality environments using HoloLens 2.

C
C++
Unity
HoloLens 2
View details

Secure Communication System

FPGA-based secure communication implementation with cryptographic primitives and hardware random number generation.

Python
DE1-SoC
Quartus Prime
View details

CPU Simulation and Cache Optimization

Architecture simulation with cache optimization for RISC-V processors to improve performance metrics.

RISC-V Assembly
Ripes
View details
Skills

What I work with

Languages

Python
C/C++
SQL
PowerShell
Bash

Frameworks & Libraries

TensorFlow
PyTorch
JAX
Scikit-Learn
Hugging Face Transformer
LangChain

Technologies & Tools

Kubernetes
Docker
OpenAI API
Hugging Face

Cloud development

Azure
AWS
GCP

Hardware

Arduino Uno
ATmega328P
BeagleBone Black
DE1-SoC
ESP32
FPGAs
HoloLens 2
Raspberry Pi Pico 2

Certifications

Microsoft Certified Azure Fundamentals (AZ-900)
Contact

Let's build something.

Feel free to reach out if you'd like to discuss potential collaborations or just want to connect.