I love building systems at the intersection of crypto and applied AI.

Solve UX & perf bottlenecks by pushing public deterministic work to a server and retain private cryptographic work client-side.

• Public-deterministic work: chain scanning, Merkle-tree reconstruction. Cacheable, shareable, identical for every user.
• Private cryptographic work: balance and note decryption that must stay client-side to preserve privacy.
• Run the public half once on a server and hours-long sync reduces to seconds. Unlocks mobile, chatbot, and agentic private payments.

On EigenLayer, a trader sits two degrees removed from the yield bearing asset, dampening any directional view a trader might want to express.

• LRTs give blended basket exposure, like an index fund without direct access to individual assets (AVS per-asset yield).
• LAT is a restaking primitive that unbundles AVS selection, per-asset selection, and operator switching into separate choices.
• This turns basket yield into a per-AVS yield curve that products like fixed and leveraged yield can price against.

Reverse-engineered the client-side defenses of 10 major chatbot apps (ChatGPT, Claude, Grok, etc.)

• Every bot defense can be distilled into a combination of 3 defenses: cookie auth, client-minted tokens, opaque request protocols.
• Bundling into a Chrome extension allows breaking environment-level defenses (fingerprinting, behavioural) for free.
• Connecting to a backend allows automated prompting at scale across real-user accounts (with their permission).