Phoenix - Your AI AssistantSenior Backend Engineer: Optimize Automation System
- Posted:
- Proposals: 54
- Remote
- #4527828
- Open for Proposals
Manchester 
Description
Experience Level: Expert
We're a Los Angeles leasing company . Our whole operation runs on 14 automations on Railway: tour confirmation texts, driver schedules, owner access texts, application intake, DocuSign contracts, invoicing, and a sync between Monday.com (our CRM) and Quo/OpenPhone (our phones). About 57,000 lines across Node.js and Python, several using the Claude API to read customer replies and write reports.
It works and it runs our day. But it was built fast, one fix at a time, and it shows:
- The same business rules are written in 5 places in two languages, and they drift apart.
- 11 separate Monday connections and 25 files talking to our phone API. Every service polls on a timer, and we're using about 3 times Monday's daily API limit.
- AI costs are high: full AI agents run for routine reports, with no caching.
- One test file across 14 services. No version history, no staging environment, and deploys come from a laptop.
- Mistakes are found by our team after customers see them.
WHAT YOU'LL DO
- Put everything in GitHub with CI, plus a staging environment that uses test boards and a test phone line.
- Build one shared library: a single client each for Monday, Quo and Gmail, with rate limits, retries and call counting, and a single rules module, used by every service.
- Replace polling with webhooks wherever the APIs allow, and cache the rest.
- Cut AI spend: use plain code where a rule will do, cache prompts, and match each task to the right model.
- Turn our written rules and our long list of past mistakes into automated tests, so a fix can't break an earlier one.
- Add a simple dashboard: what each service did, what it cost, and what failed.
- Then stay on part-time (about 5-10 hours a week) to review changes and handle incidents.
WHAT YOU WON'T DO
Redesign how we run the business. Our rules are detailed, written down and settled. Your job is to make the system follow them reliably and cheaply. You'll work alongside the AI coding assistant that wrote most of this code and knows the rules.
MUST HAVE
- 5+ years building production backends in BOTH Node.js and Python
- Webhook and event-driven design
- Heavy third-party API integration: rate limits, retries, idempotency
- Writing tests for messy business logic
- Comfort deploying containers on Railway, Render, Fly or similar
STRONG PLUS
- Monday.com GraphQL API
- OpenPhone/Quo or Twilio
- LLM cost optimization
- DocuSign or QuickBooks APIs
It works and it runs our day. But it was built fast, one fix at a time, and it shows:
- The same business rules are written in 5 places in two languages, and they drift apart.
- 11 separate Monday connections and 25 files talking to our phone API. Every service polls on a timer, and we're using about 3 times Monday's daily API limit.
- AI costs are high: full AI agents run for routine reports, with no caching.
- One test file across 14 services. No version history, no staging environment, and deploys come from a laptop.
- Mistakes are found by our team after customers see them.
WHAT YOU'LL DO
- Put everything in GitHub with CI, plus a staging environment that uses test boards and a test phone line.
- Build one shared library: a single client each for Monday, Quo and Gmail, with rate limits, retries and call counting, and a single rules module, used by every service.
- Replace polling with webhooks wherever the APIs allow, and cache the rest.
- Cut AI spend: use plain code where a rule will do, cache prompts, and match each task to the right model.
- Turn our written rules and our long list of past mistakes into automated tests, so a fix can't break an earlier one.
- Add a simple dashboard: what each service did, what it cost, and what failed.
- Then stay on part-time (about 5-10 hours a week) to review changes and handle incidents.
WHAT YOU WON'T DO
Redesign how we run the business. Our rules are detailed, written down and settled. Your job is to make the system follow them reliably and cheaply. You'll work alongside the AI coding assistant that wrote most of this code and knows the rules.
MUST HAVE
- 5+ years building production backends in BOTH Node.js and Python
- Webhook and event-driven design
- Heavy third-party API integration: rate limits, retries, idempotency
- Writing tests for messy business logic
- Comfort deploying containers on Railway, Render, Fly or similar
STRONG PLUS
- Monday.com GraphQL API
- OpenPhone/Quo or Twilio
- LLM cost optimization
- DocuSign or QuickBooks APIs
New Proposal
Login to your account and send a proposal now to get this project.
Log inClarification Board Ask a Question
-
Hi Terry,
Are the 14 automations currently in a single repository or spread across separate projects, and do you have documented business rules and a list of past failures ready for us to use when building the shared modules and regression tests?
I look forward to your reply.
Best Regards. -
Are the 14 Railway automations currently deployed as separate services/repositories, or are they already sharing any common codebase?
-
One question: do you have Monday usage stats showing which services drive the overage, and can I get read-only repo and Railway access to start the audit?