Proof of work
Projects
Not a GitHub dump — a few things that show how I think. The flagship is the meter; the rest trace the same instinct backward.
vllm-cost-meter
Objective live telemetry + effective cost-per-million-token meter for vLLM servers.
A read-only observer for running vLLM servers that ingests Prometheus metrics and surfaces live effective LLM serving cost against the operator's actual traffic.
- Reads vLLM Prometheus metrics
- Tracks throughput, request rate, TTFT, TPOT, E2E latency, prompt / generation lengths, batch state, and KV cache
- Computes live effective cost-per-million-token visibility
- Ships benchmark reference curves from the paper
- Independent, reproducible research artifact
Data, finance & SEC research
SEC EDGAR / 10-K ETL
source on requestPipelines that scrape and structure financial filings for accounting and fraud-signal research.
What it shows — Comfort with messy real-world data and the patience to model it properly.
University at Buffalo faculty scraper
source on requestData-extraction tool that collected and normalized faculty information at scale.
What it shows — Early proof of the same habit: automate the boring extraction, keep the signal.
Product experiments
Propael
source on requestJob-application tooling — ATS-friendly cover letters and application workflow.
What it shows — Product sense: build the thing people actually need, ship it, learn.
Reinforcement learning web app
source on requestInteractive app for exploring and teaching RL concepts in the browser.
What it shows — Teaching instinct — understanding something well enough to make it playable.
Creative tools
TextExtend-AI
source on requestAI-assisted document and text tooling experiment.
What it shows — Curiosity about where language models help real writing work.
Older projects are framed as curiosity and shipping, not headline claims.