Automation & AI Developers — Ongoing Project Work Across Multiple Clients

#1. What this is

We run an automation agency in Quebec, Canada. We build systems for small and mid-sized businesses that are stuck doing manually what software should be doing for them.

I'm not hiring for one project. I'm building a bench of developers I can bring in as client work comes through the door. Different clients, different industries, different problems — construction, trades, professional services, retail, clinics, logistics. What stays constant is the shape of the work: find the bottleneck, build the system, deploy it, document it, hand it over.

If you're the kind of developer who gets bored building the same funnel twelve times, this is the opposite of that.

---

#2. The kind of work that comes up

You will not do all of this. Tell me which parts are yours.

**Lead capture and response**

- Instant multi-channel response to inbound leads (SMS, email, chat, WhatsApp)

- Voice agents that answer calls, qualify, book, or escalate to a human

- Missed-call text-back, form routing, lead scoring and assignment

**Scheduling and dispatch**

- Booking systems tied to real availability, crew capacity, or travel time

- Reminder and confirmation sequences to cut no-shows

- Route and job assignment logic

**Quoting and estimating**

- Intake-to-draft-quote generation from historical pricing

- Proposal assembly and e-signature flows

- Approval routing

**Knowledge and retrieval**

- RAG pipelines over a company's own documents — contracts, price lists, SOPs, spec sheets, policy manuals

- Internal assistants that give staff accurate answers instead of "ask Marc, he knows"

- Search across systems that were never designed to talk to each other

**Document and data processing**

- Invoice, receipt, PO, and contract extraction into structured data

- Intake form parsing, OCR pipelines, validation and exception handling

- Data cleaning, deduplication, enrichment, migration between systems

**Back-office operations**

- Inbox triage and drafted replies

- Invoice reminders, collections sequences, payment links

- Inventory and reorder triggers, supplier price monitoring

- Employee onboarding checklists, timesheet and payroll prep

- Compliance and reporting deadlines that currently live in someone's head

**Retention and reputation**

- Post-job review requests and response handling

- Reactivation campaigns against dormant customer lists

- Churn signals and renewal nudges

**Visibility**

- Dashboards that show owners what's actually happening: lead volume, response times, conversion, job margins, capacity

- Alerting when something breaks or a number moves the wrong way

**Integration and infrastructure**

- Connecting CRMs, accounting, field service, scheduling, and phone systems that don't natively integrate

- Custom API work where no off-the-shelf connector exists

- Deployment, monitoring, error handling, and cost control

---

#3. What I'm looking for in a developer

- **Range over specialization.** I need people who can move from a Twilio flow to a document pipeline to a dashboard without needing a new hire each time.

- **Strong Python**, plus fluency with at least one workflow platform (n8n, Make, Zapier, GoHighLevel, Activepieces).

- **API-first thinking.** Most of this work is gluing systems together well, not training models.

- **LLM application experience**: prompt design, structured output, evaluation, and knowing when an LLM is the wrong tool.

- **Vector databases** for retrieval work — Pinecone, Qdrant, Weaviate, pgvector, whatever you prefer.

- **Telephony and messaging APIs**: Twilio, Vapi, Retell, or equivalent.

- **Deployment you can hand off.** Cloud-native, documented, maintainable by someone who isn't you. I will not accept systems only one person can touch.

- **Data handling judgment.** Client data, sometimes personal or financial. You treat credentials and PII seriously.

- **French/English is a strong plus.** Many clients operate in French and their customers write in French. Not a requirement for every project, but it decides who gets certain ones.

#4. How I work

- Each project is scoped and priced on its own — hourly or fixed price per phase, your preference once scope is locked

- I handle the client relationship, requirements, and decisions; you build

- Fast approvals, no committees

- Work comes in waves as clients sign, not on a fixed schedule — be honest about your availability

- Good work gets repeat work. That's the whole model.

---

#5. To apply, answer these

1. From the list above, which categories have you actually shipped to production? Name the specific projects.

2. Which would be new territory for you, and would you take them on anyway?

3. What's your default stack for automation work, and what would make you deviate from it?

4. Describe a system you built that someone else had to maintain afterward. How did you hand it over?

5. Tell me about an automation you were asked to build that you argued against. What did you propose instead?

6. Current availability — hours per week, time zone, and how you prefer to communicate.

Start your reply with the word **BLUEPRINT** so I know you read this rather than skimmed it.

Templated proposals get ignored. Show me something you built.

Most people trying to fix a home appliance are stuck flipping through a PDF manual or guessing which part is causing an issue. OnSense set out to replace that with an app where entering a model number instantly surfaces specs, part breakdowns, and an AI assistant trained specifically on appliance repair knowledge — not a generic chatbot.

What I worked on

As part of the mobile engineering work on the React Native (Expo) app, I built the OCR module that lets users scan an appliance's nameplate with their camera to automatically identify the model and pull it into the system — removing the need to manually type in model numbers that are often worn, tiny, or hard to read.

Stack:

React Native (Expo) for iOS and Android

OCR/camera-based scanning for nameplate recognition

AI chat integration (OnSense GPT) for appliance-specific Q&A

Backend integration with the platform's parts and repair-knowledge database

What's in the app:

Model lookup by number, with detailed specs and part breakdowns

Exploded View module — an interactive, clickable diagram of appliance parts

OnSense GPT chat assistant for error codes, repair steps, and part compatibility

Nameplate scanning via OCR for fast, accurate model identification

Parts cross-referencing against verified compatibility databases

The OCR-on-nameplate angle is the right call, those plates are half worn off and users mistype the model constantly. Accuracy there lives or dies on preprocessing before OCR, glare and low contrast on metal wrecks raw recognition, so that is where the module earns its keep.