Benjamin Faruna

Software that holds up in production.

I build across AI engineering, backend systems and blockchain: RAG pipelines, APIs and smart contracts. A chemical engineering degree taught me to care about what flows through a system, and how it is measured.

See my projects View CV GitHub LinkedIn

Projects

Document Q&A, on-chain protocols, lending and data science: shipped work and what I am building next. Filter by role.

  • DocuChat

    In progress

    A RAG system that answers questions from uploaded documents. It has role-based access, response caching, logging and observability, with pgvector as the vector store. In progress.

    • RAG
    • pgvector
    • PostgreSQL
    • RBAC
    • Caching
    • Observability

  • MedVerify NG

    In progress

    A case study for a phone app that checks medicine packs in seconds. Users photograph a pack, AI reads the label and NAFDAC number, matches it to registered records and gives a risk score, and scans feed a hotspot map for regulators. The proposal covers users, risks and how to measure success.

    • OCR
    • Image recognition
    • Risk scoring
    • Case study
  • Screenshot of HODL 1 / 2

    HODL

    A blockchain platform for keeping your assets and growing your portfolio. I co-founded it and lead protocol engineering: designing, implementing and auditing the smart contracts behind asset management and security across EVM chains.

    • Solidity
    • Foundry
    • EVM
  • Screenshot of LendBit 1 / 2

    LendBit

    A cross-chain lending protocol that lets users borrow on one blockchain using collateral from another, with lending pools and peer-to-peer lending options powered by Chainlink CCIP.

    • Solidity
    • Chainlink CCIP
    • Cross-Chain
    • DeFi
    • Hardhat
    • React
    • TypeScript
  • Screenshot of PeerLend 1 / 2

    PeerLend

    A peer-to-peer lending platform where borrowers set their own loan terms and lenders send full or partial offers. Collateral is capped at 85% of its value, Chainlink Automation handles liquidations, and Gitcoin Passport handles user verification. I wrote Solidity contracts and connected the React frontend with ethers.js.

    • Solidity
    • Chainlink
    • Foundry
    • React
    • ethers.js

  • Foodie

    A team-built food delivery app where customers browse meals from many restaurants, fill a cart and place orders. I worked on the full-stack app: a REST API for customers, restaurants, foods, carts and orders, with session login and image uploads.

    • Node
    • Express
    • MongoDB
    • Passport
    • React

  • IP Music Recommender

    A web app that recommends new songs from a track you pick, with short previews so you can build a playlist without playing full songs. Built with two teammates on the Spotify API; I worked on the database model and the API.

    • Flask
    • Spotipy
    • SQLAlchemy
    • MySQL

  • Terrorist Attack Outcome Prediction

    A team capstone from the Hamoye internship using the Global Terrorism Database (over 180,000 attacks, 1970 to 2017). We cleaned the data, engineered features and built a classification model to predict attack outcomes.

    • scikit-learn
    • Pandas
    • Classification
  • Screenshot of Warden 1 / 2

    Warden

    An AI agent with its own on-chain wallet on Base Sepolia. It calls tools on its own and pays for paywalled APIs automatically using the x402 pay-per-request protocol. Next up: on-chain spending limits, so even a prompt-injected agent cannot drain the wallet. In progress.

    • TypeScript
    • Next.js
    • viem
    • x402
    • Base Sepolia

About

From chemical engineering to software, with a detour through AI training data.

I trained as a chemical engineer and moved into software, where I now work across AI engineering, backend systems and blockchain.

At Hugo Technologies I audited the datasets used to train AI models. It taught me how much data quality decides what a model can do, and I bring that care to every system I build.

Chemical engineering taught me to care about what flows through a system, and how it is measured.

Right now I am building DocuChat, a document question-answering system with access control, caching and observability, and I co-lead protocol engineering at HODL.

Skills

The tools I reach for, grouped by discipline.

  • AI engineering

    RAG pipelines, pgvector, embeddings, caching, evaluation, scikit-learn, Pandas, NumPy.

  • Backend

    Python, Flask, Django, Node.js and Express, REST APIs, PostgreSQL, MySQL, MongoDB, role-based access.

  • Blockchain

    Solidity, Foundry, Hardhat, Starknet, ethers.js, EVM-compatible chains, contract auditing.

  • Quality and data

    Dataset auditing for AI training at Hugo Technologies, statistical analysis, Power BI, Tableau.