AI automation, custom software, and on-chain engineering - held to one standard: invariants checked, logic audited, and a human in the loop where it counts.
3,500+
tests across shipped systems
200+
smart contracts engineered
5+
AI products shipped
20+
Protocol Builds
15+
Project Shipped
[ 01 - the failure modes ]
We've seen where AI and on-chain systems actually fail in production. Every engagement starts by designing against these - not discovering them later.
Let's fix itAI pilots that never leave the pilot
Audits that count tests, not invariants
Automations that break silently
Composability risk nobody stress-tested
Chatbots that hallucinate in front of customers
A multisig that's theater, not security
[ 02 - what we do ]
Workflow and ops automation that replaces manual busywork - monitored and reliable, not a fragile no-code chain.
Python · Redis · Postgres
Conversational interfaces grounded in your actual data, not generic prompts that drift off-brand by the third reply.
Claude API · pgvector · RAG
Full-stack products built from first principles - no page-builder shortcuts, no dependencies you can't explain.
Next.js · tRPC · Postgres
Fitting AI into what you already run - CRM, support, internal tools - and telling you honestly where it doesn't belong yet.
APIs · Agents · Evals
Smart contracts and dApps engineered for composability from day one, not bolted together after launch.
Solidity · Foundry · wagmi
Lending markets, perps, vaults - on Solana and Ethereum - with invariant tests and fuzzing before a dollar touches them.
Anchor · Foundry · Groth16
iOS and Android apps from a single codebase - native feel, push notifications, offline support, and store-ready release pipelines.
React Native · Expo · Push
[ 03 - where this shows up ]
AI automation and blockchain engineering aren't industry-specific - the same rigor applies whether you're moving fiat or moving tokens.
[ 04 - how we work ]
Week 1
Scope the actual problem, map the data and systems involved, and agree on what "done" looks like before any code gets written.
Week 2–4
Built against your real data and constraints - invariant tests and fuzzing where it's on-chain, evals where it's AI.
Week 4–5
Shipped with monitoring and fallbacks from day one. No silent failures, no black box you have to take on faith.
Ongoing
Weekly check-ins, not quarterly reports - and direct access to the person who built it, no account manager in between.
[ 05 - the work ]
Everything below is code that exists, tests that pass, and in two cases, a live URL you can open right now.
lendfi/
An Aave-equivalent lending market, built from first principles - then put through its own self-audit, which surfaced multiple high-severity findings despite hundreds of passing tests.
27
contracts
670
tests
Solidity · Foundry · Chainlink
tripwire/
Real-time invariant monitoring for DeFi protocols - the tool a few of the other projects on this page could have used before they shipped.
9
packages
685
tests
tRPC · Next.js · TimescaleDB
[ 06 - get in touch ]
Email works best for scoping - say what you're trying to build, and you'll get real questions back, not a sales deck.