Case studies

What each system does, what was hard about it, and what changed because of the work.

  • 13×
    RAG platform

    I inherited a retrieval-augmented generation platform and made it faster, more complete and harder to break.

    Python, FastAPI, Postgres, pgvector, AWS

  • 5–10M
    SMS platform

    A live multi-tenant platform handling 5 to 10 million messages a day, with billing and a client portal.

    Java, Spring Boot, Jasmin, RabbitMQ, Postgres, Redis, React

  • 98%
    Compliance ingestion pipeline

    A pipeline that picks up expert-call transcripts from Google Drive, screens them for material non-public information with AI, and keeps an audit trail. I built the AWS infrastructure on my own in Terraform.

    TypeScript, Node.js, AWS Lambda, AWS SQS, API Gateway, Terraform, Supabase, Postgres, pgvector, OpenAI

  • 4
    USSD banking platform

    A production multi-telco USSD platform that lets a Nigerian bank's customers bank from any mobile phone, with no internet needed.

    Java, Spring Boot, USSD, SMPP, SOAP, REST, Redis, SQL

  • 939
    CRM sync and dashboards

    A sales dashboard fed by a live read-only CRM sync, with an AI assistant that drafts but never sends.

    Node.js, Express, MongoDB