BLUEBERRY Intelligence & Automation , home

Work

Systems built for real operational problems.

A selection of systems designed and built by our founder, Victor Kola. Each one is described as the problem, the approach and the system delivered.

We do not name clients or publish results without permission, and we do not publish numbers we cannot verify. Full client case studies will be added here as they are approved.

  • Public health DHIS2 integration In development

    Public health data assistant (DHIS2)

    Problem
    Public health teams spend a lot of time running the same indicator queries in DHIS2 and turning the results into reports.
    Approach
    A conversational layer that turns plain-language questions into structured requests to the DHIS2 Analytics API, using validated indicator definitions, with a deterministic fallback when the AI is unsure.
    System
    An analytics assistant that answers indicator questions from DHIS2 data and produces tables, charts and report-ready summaries.
    • Python
    • FastAPI
    • DHIS2 Analytics API
  • Internal operations Role-based access Built

    Secure internal knowledge assistant

    Problem
    Operational knowledge was scattered across drives and manuals, so senior staff kept answering the same policy questions.
    Approach
    A retrieval-augmented assistant with role-based access control, so each person only gets answers from documents they are allowed to see, plus routing of unresolved questions to the right team.
    System
    An internal assistant that answers from authorized documents with citations and turns questions it cannot answer into routed support tickets.
    • Python
    • FastAPI
    • LangChain
    • Role-based access control
  • Customer support Hybrid search Built

    Multi-tenant customer support platform

    Problem
    General-purpose chatbots gave confident but wrong answers and could not keep each organization’s rules and documents separate.
    Approach
    Strict tenant isolation, hybrid retrieval (keyword + dense vector search) and local re-ranking, so answers come only from the right organization’s documentation.
    System
    A support platform that turns each organization’s documentation into grounded answers and hands complex requests to people.
    • Python
    • FastAPI
    • Hybrid search
    • Re-ranking
    • Next.js
    • TypeScript
  • Governance & operations Local transcription Built

    Privacy-first meeting minutes

    Problem
    Organizations wanted structured minutes from confidential meetings without sending raw recordings to a cloud service.
    Approach
    Transcribe audio locally, remove personal information on the device, and only then use a language model to structure the redacted text.
    System
    A meeting tool that produces minutes, decisions and action owners while keeping raw audio and unredacted transcripts on the machine.
    • Python
    • Whisper (local)
    • spaCy
    • Gemini API
  • Education LLM-as-judge checks Built

    Curriculum-aligned lesson planning assistant

    Problem
    Teachers spend a lot of time writing lesson plans and assessment material that match the curriculum.
    Approach
    Generate lesson plans and assessments from curriculum data, then run an automated review step (LLM-as-judge) that checks each output against the syllabus before it is shown.
    System
    A lesson-planning platform for secondary teachers that drafts curriculum-aligned plans and assessments for teacher review.
    • Python
    • FastAPI
    • Supabase
    • LLM-as-judge evaluation
  • Education Optimization Built

    Automated timetable scheduler

    Problem
    Building school timetables by hand took administrators a long time and still produced room and teacher clashes.
    Approach
    A genetic algorithm that searches many candidate timetables against hard constraints (no clashes) and soft preferences (balanced workloads).
    System
    A stateless scheduling API that returns clash-free timetables and makes better use of rooms and teachers.
    • Python
    • FastAPI
    • Genetic algorithms
  • HR & recruitment Document intelligence Built

    Structured CV screening tool

    Problem
    Small hiring teams receiving hundreds of CVs struggled to review them quickly and consistently.
    Approach
    Extract qualifications and experience from each CV and score every candidate against the same written job criteria using embeddings and a language model.
    System
    A screening tool that ranks applicants on defined criteria and writes summaries that support, rather than replace, the hiring manager’s judgment.
    • Python
    • FastAPI
    • LangChain
    • Embeddings

Facing a similar problem?

Tell us about it and we will tell you honestly whether AI is the right tool. We reply within two business days.

Discuss a project ada@blueberryia.com