Adam Herman
Commercial Counsel · Payments · AI · San Francisco
Dual-qualified U.S. and Australian attorney. In-house commercial counsel at Block — negotiating enterprise payments and agentic AI partnerships across Square, Cash App, and Afterpay.
Work Experience
Lead commercial counsel for Block's enterprise merchants and agentic AI partnerships across Square, Cash App, and Afterpay — billions in annual GPV contracted; also built the team's legal ops automation stack.
Managed merchant contracting for Cash App and Afterpay across North America; promoted from Contracts Associate within first year.
Managed BNPL merchant contracts through hypergrowth; served as in-house sell-side associate on Block's $30B acquisition of Afterpay, then acqui-hired to San Francisco.
Financial services regulation and M&A litigation across the Australian financial services sector. Fixed-term engagement.
Negotiated partnership agreements for the Office of External Relations with major telecoms, Big Tech, national health systems, and venture-backed startups.
Education
| University of Sydney | LLB · BA (Media & Comms; Government & International Relations) |
| UC Berkeley | Commercial Contract Fundamentals, Executive Education |
| London School of Economics | Global Communications, Summer School |
| Peking University | Mandarin, Winter School |
Credentials
Admitted
- New York State Bar
- Supreme Court of NSW, Australia
- High Court of Australia
Talk to me about
Articles
Canva surge a strong signal to next generation ↗ AFR The University of Sydney Law School on LinkedIn ↗ LinkedIn The 'limitless' opportunities with a law degree ↗ Sydney LawProjects
Street parking in SF is its own sport. My street has a 2-hour limit the city enforces unpredictably, and I got tired of circling the block hoping my spot was free. So I built a tool to answer one question: is there parking in front of my apartment right now?
A CCTV camera watches the spot 24/7 from my window. A local vision-language model checks the feed every 5 minutes and returns a simple verdict. A web dashboard — and an iPhone lock screen widget — surfaces the answer wherever I am. Everything runs on a Mac Mini I already owned, with no cloud APIs and no ongoing cost.
Capture → A cron job fires every 5 minutes, running a Python script that uses ffmpeg to grab a single frame from the camera's RTSP stream, then resizes it to 960px wide to reduce inference payload.
Local LLM inference → The frame is base64-encoded and sent to LM Studio running locally, which serves Gemma 4 (a vision-language model) via an OpenAI-compatible API. The prompt is engineered around a physical landmark in the frame — a street pole — giving the model an unambiguous spatial boundary to reason about. Temperature is set to 0 for fully deterministic output. The model replies with exactly ANSWER: FREE or ANSWER: OCCUPIED.
Server → A pure-Python HTTP server (no frameworks) runs continuously on port 8181. It serves the dashboard and a set of JSON API endpoints for status, the live camera image, snapshot history, and occupancy stats. Results are written to SQLite; the latest frame is stored for the live dashboard image.
Dashboard → A self-contained HTML app shows a live status badge, the latest camera snapshot, occupancy stats (today / this week / all time), a bar chart of historical occupancy, and a scrollable snapshot grid with colour-coded thumbnails. Password-protected with PBKDF2-hashed credentials.