Currently leading R&D efforts at A10 Networks to build AI-driven firewalls that protect generative AI applications against emerging cyber threats. I design and deploy advanced LLM guardrails that ensure safe and reliable AI usage across enterprise environments.
Beyond my role, I’m deeply passionate about democratizing access to artificial intelligence. I've built LLMs from scratch and developed an open-source platform that enables anyone to run large language models locally — with both backend and frontend fully integrated, requiring no internet connection. My work is driven by a belief in AI’s transformative potential when it is open, safe, and accessible.
AI/ML
70%
Cloud
50%
System Designing
60%
what i do.
Machine / Deep Learning
Learn & Build LLM Powered Applications
Exploring and researching emerging LLM architectures and advancements.
Fine-tuning LLMs using state-of-the-art techniques like SFT, QLoRA, and RL.
Deploying LLMs efficiently with vLLM, Docker, Helm, and Kubernetes for scalable production use.
recent work.
Here's some of my recent work
all
llms
GPTs From Scratch
LLMs for Everyone - Fully Offline
Project Overview
GPTs From Scratch
An open source GPT-based model implementation using PyTorch, such as Llama & Mistral. Aim is to gain practical expertise in training and deploying these models to production, as well as a theoretical understanding of the underlying working principles.
LLMs are reshaping software—but running them locally is still way too complex. Most tools rely on paid APIs, fragile setups, or a tangle of frontend/backend dependencies. Me and my friend wanted to change that. So we built an open-source project that makes running LLMs locally simple, free, and offline-first. Key features include:
Run any Hugging Face model locally — no internet after setup
Chat & Completions playground
One tool, no stitching required — backend and frontend in one place
Built for accessibility. Powered by open-source. No cloud lock-in.