I take any problem
from a blank page
to a product people trust.

From a prop-trading idea to credit infrastructure any DeFi market can plug into.

Vanna began in November 2023 as a prop-trading protocol: lenders would fund traders who had the skill but not the capital. Building it, testing it with early users and taking it to investors and trading firms showed us the real gap sat underneath the trade. On-chain credit is either overcollateralised, returning less than you post, or trapped in the venue that issued it, so a hedge on one venue cannot support a position on another and most capital sits idle as excess margin. Prime brokers solved this decades ago by financing a client’s whole book from one account. Rebuilding that on-chain is first a risk problem: credit that exceeds collateral and moves across markets is only safe if the engine can see, price and close every position it has financed. With my cofounder I worked out that model, and I shaped the product around it: how traders, businesses, institutions and agents reach the same credit, and how each of them sees the risk before they act.

When the same collateral can back positions in many markets, more of it is put to work, which adds liquidity and volume across DeFi. The current protocol runs on Stellar and Solana and is being prepared for a limited mainnet launch.

I sold, scoped and designed this programme, and now run its delivery on site; the first module is in user acceptance testing.

Most enterprise AI stalls between the demo and the ERP: leadership wants a return, IT wants to know exactly what the model can touch, and the people at each desk need it to fit the way they already work. This programme was built to clear all three. I led it through my company, Zonymous Labs, from the first workshop, and now lead delivery on site. It began with a research file and a questionnaire for each company, then workshops with leadership and the department heads, and a catalogue of candidate use cases narrowed to four modules: order-to-contract, procure-to-pay, accounts and finance, and export. The business case was sold on ranges and peer evidence, with one module first as a proof of concept, and the security answer went into the proposal and then the contract. Most of the rules lived only in people’s heads, so I sat with each desk, recorded how the work is done and wrote those rules down for the first time; deterministic code now applies them, and agents handle what varies from one document to the next. The ERP is touched only through one governed path. I am now turning the same platform into an agentic OS for manufacturing that can be repeated at any plant running an ERP.

The film manufacturer’s domestic order module is in user acceptance testing, tested end to end against the client’s ERP; procure-to-pay is in build, and export and finance follow. Results will be measured against manual entry once each module is live; volumes stay under the client’s NDA.

Gold that works like money, on DeFi rails its users never have to see.

People trust gold as savings, but it cannot pay for a coffee, cross a border or back a loan in seconds. Tokenised gold fixed the plumbing, but the services around it sit scattered across DeFi protocols and chains, built for people who already live in crypto, while most people who save in gold have never opened a crypto wallet. Auri started from that gap: a gold neobank where someone signs in with an email or a passkey, sees their money in their own currency and never meets a seed phrase, a gas fee or a DeFi term. Underneath, tokenised gold, a lending protocol, swap routing and smart accounts do the work. I conceived the product and researched the customer, the competition, the business model and Canadian regulation first; then I chose and composed the rails, designed every surface from web to mobile, and built the working prototype, which I validated with my CTO. Tokenised gold is issued at scale but barely reaches ordinary savers, and Auri is designed to reach them.

Buy, sell, borrow and repay run on Arbitrum mainnet in private testing; gifting and Auto-Invest are prototyped; card and bank funding, remittance and the spending card are designed and next.

From outside the company, I scoped Perflo’s on-chain money layer for its first build.

The founder came with an idea, agents that hold and spend real money, and open questions wherever crypto touched it. I began with discovery: which flows needed stablecoins or on-chain rails at all, and what an agent should be allowed to do with them. From there I framed the on-chain layer as an account an agent spends from, under limits the owner sets: caps, allowlists and approvals checked before money moves, which the agent can never widen. We scoped what came first and what waited, so the crypto side had a path from idea to first build.

I wrote the circuits, the browser proofs and the on-chain programs myself.

After the ZK certification I wanted to prove the thesis in code: a borrower should be able to show they are collateralised without revealing what they hold. ShieldLend does that twice. On EVM, Circom circuits produce Groth16 proofs in the browser, with Pedersen commitments, Merkle membership and nullifiers, and a range proof that the collateral ratio clears the threshold while the amounts stay hidden; verification runs through zkVerify at roughly a tenth of the L1 cost. On Solana, three Anchor programs verify Groth16 on-chain, with privacy layered across the whole transaction lifecycle.

I designed ZTips end to end and framed how its privacy works, from the tipster’s proof to the payout.

Anonymous tip lines fail in three ways: nobody can tell a real tip from a malicious one, a careless system can trace the tipster, and giving information can put a tipster’s life, family or money at risk. ZTips answers all three with zero-knowledge proofs, credibility ratings and escrow. Our team of three built the prototype at ZK Hack Montréal in August 2024. It began as a whitepaper and an Ethereum proof of concept I wrote in 2018 around bounties, anonymous replies and escrow; we set it aside while zero-knowledge tooling matured, and the 2024 model added proofs, ratings and rewards released in stages.

I cofounded Koinfox and ran its product and growth until it was acquired in 2020.

Active crypto traders kept accounts on several exchanges and switched between screens to trade them. Koinfox put those exchanges behind one trading terminal, built on the exchanges’ own infrastructure, with execution across them and the workflows professional traders expect. I led the product team and the growth work, and Koinfox became Southeast Asia’s first Binance Broker Partner. It is also where I learned what leveraged traders need: most of what I know about margin, liquidations and behaviour under stress comes from sitting with those users.

In 2017 I led EthLend’s entry into India, when Aave was still a peer-to-peer lending marketplace.

EthLend matched borrowers and lenders peer to peer on Ethereum, years before lending became the core of DeFi. I joined in June 2017, in the project’s first months, and spent the next eight months taking it to India’s colleges, institutes, government programmes and hackathons. I contributed to the whitepaper and coordinated the smart-contract audits behind the token sale. Hacks had already started across the industry, and keeping that sale clean is the part I am proudest of.

I have taught blockchain since 2017, judge hackathons and back early founders.

Since 2017 I have taught blockchain as a financial system and a technical one: government institutes ICAI and NPTI across six cities including Dubai, engineers at IITs, Delhi University and PEC, and colleges under an Andhra Pradesh Government initiative. I judge technical hackathons, lately ETHGlobal Delhi and ETHDenver, and have advised or backed ten-plus startups, Aconomy and KoinX among them. Cut The Noise, my YouTube channel, brings the same thinking to live markets: short, plain breakdowns of crypto, macro and trading, and one rule: sometimes not trading is also a trade.