# Chitral Patil > Chitral Patil is an AI infrastructure engineer, independent researcher, open-source builder, and Hindustani classical vocalist working on LLM inference economics and vLLM cost telemetry. Spine: Discipline under uncertainty. ## About Chitral Patil is an AI infrastructure engineer/researcher and Hindustani classical vocalist. His recent work focuses on LLM inference economics: measuring the real cost of serving open-weight models under production-like constraints rather than relying on static token calculators. He created vllm-cost-meter, an open-source tool for live vLLM cost telemetry, as a companion to his arXiv paper. He has also worked across data science and software systems, and has a long-running music practice rooted in Hindustani classical vocals. ## Three threads - Systems: LLM serving, vLLM, Prometheus telemetry, GPU economics, and the cost of latency. - Sound: Hindustani classical vocals — riyaz, raag, and the long discipline of listening. - Product: Tools that solve a painful real problem, shipped and written about in the open. ## Research Beyond Per-Token Pricing: A Concurrency-Aware Methodology for LLM Infrastructure Cost Estimation Public LLM cost calculators often reduce serving economics to a static token price or assumed utilization. This work studies how request rate, concurrency, latency SLOs, hardware, model architecture, and quantization interact to change the real effective cost of self-hosted LLM inference. Key findings: - Token price is not serving cost. - Load and concurrency shape the economics. - OpenAI-compatible is not cost-compatible. ## Featured project vllm-cost-meter — Objective live telemetry + effective cost-per-million-token meter for vLLM servers. https://github.com/pChitral/vllm-cost-meter ## Projects - vllm-cost-meter: Live cost-per-million-token meter that reads Prometheus telemetry from a running vLLM server. (https://github.com/pChitral/vllm-cost-meter) - SEC EDGAR / 10-K ETL: Open-source pipelines that scrape and structure SEC 10-K filings for accounting and fraud-signal research. (https://github.com/pChitral/ETL-SEC-EDGAR-10-k-Filings) - University at Buffalo faculty scraper: Data-extraction tool that collected and normalized faculty information at scale. ## Writing - Your router does not know your evals (https://chitralpatil.com/writing/your-router-does-not-know-your-evals): Model routing is an evaluation problem. A generic router cannot know where smaller models are good enough for your workload, but a router trained on your own judgments can. - The difference has to click (https://chitralpatil.com/writing/the-difference-has-to-click): What making cover songs under unreasonable deadlines taught me about invisible effort, legible differentiation, and learning to lead the orchestra instead of playing every instrument. - The unlosable thing (https://chitralpatil.com/writing/the-unlosable-thing): Why I do the expensive version of things: the proof you can rebuild it is the only asset worth holding, the trick is controlling where you collect your dopamine, and slightly-above-average is the flywheel's ignition. - OpenAI-compatible is not cost-compatible (https://chitralpatil.com/writing/openai-compatible-is-not-cost-compatible): A matching API surface tells you nothing about what a token actually costs to serve. Two endpoints can speak the same protocol and live in different economic universes. - Meter beats calculator (https://chitralpatil.com/writing/meter-beats-calculator): Calculators predict a best case you will rarely hit. A meter reads the truth off live telemetry. Why I built the meter instead of another spreadsheet. - Riyaz and research (https://chitralpatil.com/writing/riyaz-and-research): What years of Hindustani classical practice taught me about systems: improvisation inside constraints, repetition without boredom, and staying with complexity. ## Links - Website: https://chitralpatil.com - GitHub: https://github.com/pChitral - Google Scholar: https://scholar.google.com/citations?user=WU6A-FAAAAAJ&hl=en&oi=sra - arXiv paper: https://arxiv.org/abs/2606.11690 - LinkedIn: https://www.linkedin.com/in/chitralpatil/ - YouTube: https://www.youtube.com/c/ChitralPatil/videos - Instagram: https://www.instagram.com/chitral_patil/ - X / Twitter: https://x.com/ChitralPatil - Facebook: https://www.facebook.com/chitral.patil/ - LeetCode: https://leetcode.com/u/chitralpatil/ - Email: mailto:cpatil.research@gmail.com ## Disclaimer Research and open-source work shown here are independent unless otherwise stated and are not affiliated with, endorsed by, or reflective of the views of any employer.