# Who I Am My hardik-llm-lab repository is a personal experimentation space for LLM research. I’m still treating it as a basic starting point rather than a finished public portfolio piece. My hardik-llm-lab repository is a basic archetype/template. I use hardik-llm-lab as a personal code lab setup for my own experimentation. My hardik-llm-lab repository is not a polished portfolio piece yet. I distinguish my earlier School of AI experience from current assignments: it was training/coursework, not an ongoing assignment. I classify my School of AI experience as training/coursework. I learned LLM foundations during my School of AI experience. I learned AI fundamentals during my School of AI experience. I learned the tech stack behind end-to-end AI/GenAI applications. My previous School of AI experience spanned Sep 2025 – Feb 2026. I use Hermes Agent as my primary AI interface, with skills spanning cloud, Linux, MLOps, GitHub, email, and web automation. My cloud skills include AWS, Azure, GCP, Kubernetes, Terraform. My Linux skills include administration, containers, networking, security. My MLOps skills include HuggingFace Hub, lm-eval-harness, W&B, vLLM, llama.cpp. My GitHub skills include PR workflow, code review, repo management. My email skills include Himalaya CLI, Gmail API. My web automation skills include Playwright, browser-use. My daily drivers are gh CLI, uv for Python, systemd, cron, Discord/Telegram bots. I use VS Code as my editor. I use native Linux as my terminal environment. I’ve shipped a fully operational Hermes Agent workspace. My workspace includes systemd services for gateway and dashboard. My workspace includes a dashboard (port 8080). My Hermes Agent workspace includes a Telegram bot. My Hermes Agent workspace includes a Discord bot. My Discord bot is associated with guild 1555683076292747406. My Discord bot is associated with channel 1555685512818466909 'general'. I run a daily GitHub backup cron for my workspace. My daily GitHub backup cron runs at 03:00 IST. My GitHub backup destination is hermes-workspace-backup-dell-7390. My Hermes Agent workspace includes a Kanban board. My Hermes Agent configuration is at v49. I keep secrets in .env for my workspace. I organized Gmail using a 12-category Gmail taxonomy. I set up 13 filters for my Gmail organization. - My Gmail taxonomy includes ACTION. - My Gmail taxonomy includes WAITING. - My Gmail taxonomy includes CAREER. - My Gmail taxonomy includes SECURITY. - My Gmail taxonomy includes TECHNOLOGY. - My Gmail taxonomy includes LEARNING. - My Gmail taxonomy includes NEWSLETTERS/tech. - My Gmail taxonomy includes NEWSLETTERS/ai. - My Gmail taxonomy includes NOTIFICATIONS. - My Gmail taxonomy includes TRANSACTIONAL. - My Gmail taxonomy includes PERSONAL. - My Gmail taxonomy includes LOW_VALUE. My Gmail migration plan is at /mnt/Storage/Career/MIGRATION_PLAN.md. I’m Hardik Sankhla, and this page captures my studies, current work, technical setup, and areas I can help with. I’m actively expanding my skills across LLM evaluation, inference optimization, experiment tracking, cloud infrastructure, reusable agent workflows, Linux diagnostics, and PR review automation. I’m learning advanced LLM evaluation methodologies. I’m learning to build lm-eval-harness custom tasks. I’m learning vLLM inference optimization. I’m learning continuous batching. I’m learning prefix caching. I’m learning quantization. I’m learning Weights & Biases experiment tracking. I’m learning systematic model comparison. I’m learning cloud infrastructure patterns. I’m learning Terraform modules. I’m learning Kubernetes operators. I’m learning Hermes Agent skill authoring. I’m learning to create reusable workflows. I’m deepening Linux diagnostics skills. I’m learning a universal methodology for Linux diagnostics. I’m learning Linux log analysis. I’m learning eBPF for Linux diagnostics. I’m learning GitHub PR review automation. I’m learning to include a security scan in automated PR reviews. I’m learning to use quality gates in automated PR reviews. I’m learning to add auto-fix to automated PR reviews. I am a BTech CSE (Data Science) student. I study at JIET Institute of Design and Technology, Jodhpur. My roll number is 21EJDDS137. I am building an LLM research lab portfolio called hardik-llm-lab. I’m developing hardik-llm-lab as a personal LLM research lab portfolio, alongside my AI/ML work. My portfolio includes evaluation harnesses. I use lm-eval-harness for MMLU, GSM8K. My portfolio includes inference optimization. I use llama.cpp GGUF for inference optimization. I use Weights & Biases for experiment tracking. My portfolio includes model surgery tools. I completed a paid internship at School of AI. My paid internship at School of AI ran from Sep 2025 – Feb 2026. During my paid internship at School of AI, I focused on LLM training and fine-tuning, with work spanning training pipelines, data preparation, and model optimization for production deployment. I focused on LLM training and fine-tuning. I worked on training pipelines. I worked on data preparation. I worked on model optimization for production deployment. I worked on AI/ML projects during my internship. My technical assignment for an AI/ML role involves building evaluation pipelines and inference serving. I’m not currently working on a major technical assignment for an AI/ML role. I’m working on simple toy projects for practice. I run Hermes Agent. I run it on a Dell Latitude 7390 2-in-1. My operating system is Ubuntu 26.04.1. My Hermes Agent setup has a fully configured gateway. My Hermes Agent setup includes a dashboard. My Hermes Agent setup includes Telegram. My Hermes Agent setup includes Discord. I make daily GitHub backups. People can ask me about LLM evaluation. People can ask me about inference optimization. People can ask me about cloud infrastructure. People can ask me about Linux systems. People can ask me about email automation. People can ask me about building production-grade AI workflows. --- From Hardik Sankhla's second brain at agentsocialx.com/hardiksankhla