SparkAssist: Field-Service AI For Electricians

One-Line

SparkAssist is a Telegram-based AI assistant for electricians that turns rough site notes and voice notes into structured job data, compliance documents, quotes, invoices, RAMS, and EICR PDF reports.

Why It Matters

Electricians often work from messy field context: shorthand notes, spoken observations, circuit details, customer updates, photos, and follow-up tasks. SparkAssist explores the same deployment pattern as TeachClaw in a different vertical: capture the workflow where the user already is, extract structured data from messy input, and generate documents that fit the job.

What I Built

• A chat-native Telegram workflow for rough notes and voice notes.
• Voice-note transcription path for hands-free field input.
• Structured extraction for electrician shorthand and job details.
• EICR generation pipeline with extraction, smart defaults, and PDF output.
• Workflow coverage across EICR, Electrical Installation Certificates, Minor Works Certificates, quotes/invoices, job sheets, customer messages, RAMS, and BS7671 reference support.
• Domain-specific reference support for common compliance details such as observation codes, test values, circuit data, and safety documentation patterns.

Deployment Angle

SparkAssist is useful in the AI Deployment / FDE story because it shows repeatability across domains. TeachClaw handles teacher workflows; SparkAssist handles electrician workflows. Both require:

• translating messy user input into structured data;
• designing task-specific AI routes instead of one generic chatbot;
• generating deterministic artifacts from model outputs;
• keeping the interface close to the user’s natural workflow;
• building around real operational constraints rather than demo-only prompts.

Technical Shape

• Interface: Telegram chat.
• Input: text notes and voice-note transcription.
• AI layer: domain extraction, routing, and document drafting.
• Document layer: generated certificates, reports, quotes, invoices, RAMS, and customer communications.
• Reliability pattern: structured extraction plus defaults before PDF/document generation.

Best Interview Angle

SparkAssist is the proof that the TeachClaw pattern is not only an education idea. The underlying engineering instinct is broader: identify a document-heavy professional workflow, capture messy real-world input, structure it safely, and ship useful artifacts back through a channel the user already understands.