Currently
Highlights:
- Fine-tuning LLMs (Unsloth & prompt engineering) to generate structured XML outputs.
- Built APIs to connect GUI ↔ Python backends; automated manual validation with AutoHotkey.
- Architected PLC → XML converter (150+ test cases) & fuzzed IoT devices for hardening.
Skills
Selected projects
PLC → PLCopen XML converter
Algorithm + 150+ testcases
LLM-based XML generator
Fine-tuned models for reliable structured output
Automation Suite
700+ manual testcases automated; Python library for tests
IoT Fuzzing
Fuzz scripts using Boofuzz to improve device resilience
Console — recent dev log
➜ booting local dev...
➜ loading LLM weights… done
➜ running 150 validation cases: 150/150 ✅
➜ packaging CLI → .exe: success
➜ pushing prototype to test bench…
➜ loading LLM weights… done
➜ running 150 validation cases: 150/150 ✅
➜ packaging CLI → .exe: success
➜ pushing prototype to test bench…
Persona
Equal parts curious tinkerer and methodical engineer. I prefer solid experiments, clear metrics, and code that survives real users. When not debugging an IoT stack I write little scripts to make life slightly less manual.
Current focus: LLM applications for industrial automation, robust test automation, and edge device reliability.
Contact: varunsv29102001@gmail.com · LinkedIn