Vadodara, India
Sumit VekariyaEngineer, verifiable systems & evaluation@sarkazein
Eight years shipping software, five in blockchain. I build production smart contracts and full-stack products, with a focus on account abstraction, verifiable identity, and zero-knowledge patterns. I’ve merged 45+ PRs into public docs, client codebases, and security libraries across the ecosystem.
I’m deepening work in probabilistic ML and technical AI safety, especially reliable evaluation, building on the same instincts I use for auditability and adversarial thinking in open networks.
Certified yoga instructor · Mentor at EthIndia & EthSingapore (2023–2025).
⌘ K / Ctrl K— all sections
Experience
- BounTEA: turn GitHub issues into funded on-chain bounties (escrow, lifecycle contracts, and GPG-backed verification).
- Shipped ERC-6900 GPG validation, TEA token allowance helpers, off-chain bounty storage, payouts to GPG-linked wallets, and a GitHub-authenticated dashboard.
- Evaluation research: impact-evaluation design for open-source work (audit-friendly metrics and provenance, fed BounTEA / IERR); presented at Protocol Labs’ Impact Evaluators retreat.
- Sacred Logos: DAO social layer (token gating, chat, notifications via Ceramic, Farcaster, Web3Auth, Supabase), plus ZK voting so holders can signal without doxxing wallets.
- Sacred Finance: privacy-preserving DeFi (ZK proofs, mixers, privacy scoring, and a custom relayer stack).
- Developer relations at Ethereum events; shipped integrations with IPFS, Orbis, and decentralized identity.
- Codemarket — Founding full stack developer
- Co-founded a live skill marketplace with real-time audio/video mentoring; owned the full stack from MVP to scale.
- Ran hiring and community meetups; established remote-first agile practices for delivery.
- Led ADMTC, an education platform that reached 150+ institutions and 5,000+ students; grew engineering from 1 to 5.
- Introduced CI/CD, automated tests, and architecture reviews so releases stayed safe as load grew.
Projects
- Zkred Agent ID
EIP-8004 · DIDs · TS SDK · x402 · 2025
Lets integrators register agents with DIDs on EIP-8004 using a TypeScript SDK (no Solidity required for the happy path), plus an x402 payment server (USDC / EIP-712).
- Zkred: Proof of Humanity
ZK · PrivadoID · SSI · 2024 to present
Proves personhood with ZK and short-lived (15-day) credentials so you verify once and retain less raw PII.
- ZKAccess
Next.js · account aggregator · ZK · 2024
Selective-disclosure identity checks aligned with India’s DPDP Act; shortlisted at the Sahamati Hackathon.
- TruLens
Noir · Starknet · IPFS · 2024
Camera pipeline that anchors media provenance with ZK; shortlisted at Noirhack 2025.
- Sol2Seq
Rust · Solidity · CLI · 2024
Rust CLI: point it at Solidity contracts, get sequence diagrams your team can paste into docs or reviews.
- Smart Contract Humanizer
Solidity · NLP · 2023
Reads an ABI, lets you call contracts from plain language, and returns explanations non-Web3 teammates can follow.
- De Keeper
IPFS · Chrome extension · 2022
Password vault you control, stored on IPFS; won an IPFS/Filecoin bounty track at a hackathon.
- OpenBooks
Dart · Solana · on-chain library
Peer-to-peer library of books on Solana: catalog and flows in Dart, on-chain where it matters.
AI & research
Research interests sit at identity, privacy-preserving coordination, and how we evaluate contributions in open ecosystems: the same problems that show up when we ask whether AI systems are measured honestly and held accountable.
Research themes
- Identity and accountability for autonomous and human actors in open networks.
- Privacy-preserving coordination (ZK, verifiable computation) as a template for trustworthy pipelines.
- Impact evaluation and attestations in open source, with a cryptographic framing toward evaluation reliability.
What I’m doing now
- Self-directed study toward probabilistic ML and technical AI safety (evaluation, calibration, robustness).
- Ongoing technical writing on safety vocabulary, discovery tooling, and cognitive security (see Writing below).
- Day-job thread: audit-friendly metrics and provenance for open-source contributions (Tea.xyz / BounTEA).
AI-adjacent projects
- Zkred Agent ID
Verifiable on-chain identity for agents: accountability primitives relevant to AI agent deployments.
- TruLens
Media provenance with ZK—authenticity and tamper-evidence adjacent to generative-media questions.
Writing
Filter by topic — each piece can appear under more than one tag.
- The friction deficit: how AI discovery tools are engineering consensus into the scientific method2026
- AI safety glossary: 35 terms I had to learn the hard way2026
- What Chapter 1 of Harvard's ML Systems Textbook Taught Me About AI And Why I Was Wrong2026
- The abandoned identity bug: why AI-driven science cannot scale without zero-knowledge agent attestation2026
- Cognitive security for captive minds2026
- Standardizing impact evaluation in open source: the BounTEA approach2025
- Why you don't see CAPTCHAs on your iPhone anymore2024
- Mind: An endlessly running infinite loop2024
- Too much to process: the modern epidemic of information overload2024
- The Finternet explained: a new frontier for interconnected financial systems2024
Residencies
- Impact Evaluators Research RetreatIceland · 2025
IERR · invited participant
In-residence work on verifiable impact evaluation for open source—direction that fed the BounTEA paper (listed at IERR 2025).
- Cypherpunk CampBerlin
Invited participant
Privacy- and cryptography-focused builders’ program—secure comms, decentralized identity, and verifiable systems.
Open source
I treat upstream work as part of the job: 45+ merged PRs that tightened docs, security, and day-to-day developer experience—mostly across Ethereum ecosystems (Ethereum.org, clients, security libraries). Product work spans EVM and beyond (e.g. Solana).
- Alchemy (ERC-6900) — GPG validation; factory / verifier libs
- Protocol Labs (Lotus) — StateSectorCmd DealIDs
- Ethereum.org — Docs and developer guides
- Semaphore — Privacy and documentation
- Nethermind — Client tooling and templates
- OpenZeppelin — Security libraries, Stylus contracts, test helpers
ETHGlobal New Delhi 2024 (multiple wins), IPFS/Filecoin bounty 2022; shortlists: Noirhack 2025, Sahamati.
Technical skills
- Languages: Solidity (5y), TypeScript (8y), JavaScript (8y), Rust (2y), Python
- ML & AI safety focus: evaluation design, benchmark interpretation, reliability framing, interpretability concepts, threat modeling
- Math (applied): probability, statistics, linear algebra
- Coursework (B.Tech): Artificial Intelligence, numerical & statistical methods, data mining, algorithms, theory of computation
- Blockchain: Ethereum & EVM (primary), Solana, Web3.js, Ethers.js, Hardhat, Foundry, OpenZeppelin, Wagmi, Viem
- ZK & privacy: Noir, Circom, PrivadoID, Semaphore, zk-SNARKs
- Protocols: ERC-20/721/1155, EIP-8004, EIP-6900, EIP-4337, IPFS
- Full stack: React, Next.js, Node.js, NestJS, Postgres, MongoDB, Docker, AWS, CI/CD
Education & certifications
- B.Tech computer science — Marwadi Education Foundation (2013–2017)
- Protocol Labs Dev Guild (2025)
- Encode Club Solidity Bootcamp (2024)
- Certified yoga instructor
Key achievements
- Shipped EIP-8004 extensions for DIDs so autonomous agents can register on-chain, live on multiple chains.
- Built the ERC-6900 GPG validation module (fork: gpg-modular-account), PR #334.
- Contributed Lotus StateSectorCmd so DealIDs surface with backward compatibility, PR #13140.
- ETHGlobal New Delhi 2024: Zkred Agent ID placed 2nd (Hedera AI) and 4th (Polygon Agentic Payments).
- 45+ merged PRs to Ethereum.org, Semaphore, Nethermind, OpenZeppelin: docs, security, and developer experience.
- Essays on evaluation quality, AI safety concepts, and trustworthy infrastructure (Medium / Substack).