Srikar Vardhan
Machine learning engineer & researcher
I build ML systems from the research up. Published work on LLM interpretability and compression, and apps that people actually download, like a desktop companion with 60+ stars and a speech-to-speech pipeline with voice cloning. Currently an AI/ML Research Engineer at Petavue in Chennai.
6 peer-reviewed papers
Published at AACL & COLING 2025
90+ GitHub stars across projects
B.Tech CSE, NIT Silchar (2026)
Featured work
MAX
62 ★A pixel-art desktop companion for Windows with 71 features across 8 categories. 150+ downloads on the Microsoft Store. Free, offline, zero data.
Get it on the Microsoft Store →
Speech to Speech
31 ★Translate a video into another language in a cloned voice, with the lips re-synced to match.
View on GitHub →
CircuitTier
paperFirst-author research: compress an LLM 3.84x while keeping about 99% of its task accuracy.
Read the research →
Need something like this built?
I take on select freelance work: custom desktop apps, voice pipelines, RAG assistants, or making a model smaller and cheaper to run. Tell me the problem and I will bring both the research and the engineering.