Data pipelines and workflow automation
Workflow automation from Mytrya means moving data between systems on a schedule or on an event, with the failure cases designed before the happy path. Code where the logic matters, n8n or Make where a visual workflow is genuinely simpler for your team to maintain.
The pipelines I run today pull market data from an undocumented API at a measured 1 request per second with a scraper fallback, mine a community forum against a keyword index and export scored leads to CSV, and re-run cheaply because every step deduplicates on write.
A good fit when
- Data is copied between two systems by a person, on a schedule, and it goes wrong when they're away.
- A source you depend on has no stable API and you need it to fail safely rather than silently.
- You use n8n, Make or Zapier already and the workflows have grown past what a diagram can hold.
Not the right call when
- The process changes every week. Automate it after it settles.
- A one-off migration. I can help, but it's a script, not a system.
What gets built
6 parts
- Scrapers and API clients with measured rate limits, shared pacing, and a fallback path
- Deduplication on write so any run can be repeated without side effects
- Scoring and filtering rules written down and versioned, not buried in a prompt
- Scheduled jobs on Vercel Cron, Cloudflare Cron Triggers or GitHub Actions
- n8n or Make workflows when the team will maintain them, with the tricky nodes replaced by small code functions
- Alerts to Slack or email when a source changes shape
How it's kept safe
- An unexpected response shape is treated as an outage. The pipeline stops and tells someone. It does not guess at a malformed row.
- Rate limits are measured against the real source, not assumed, and the numbers are in the code with the date they were measured.
- Every job is idempotent, so the fix for a bad run is to run it again.
Typical stack
- TypeScript or Python
- Postgres (Neon) or SQLite for small tools
- n8n self-hosted or cloud, Make, Zapier where the client already uses them
- Vercel Cron, Cloudflare Workers, GitHub Actions
Built and running
2 case studies
- NEPSE CopilotA multi-account investing copilot for the Nepal Stock Exchange that grades its own callsNext.js 16 · Neon Postgres · AI SDK v7 · Gemini Live · TelegramA multi-account investing copilot for the Nepal Stock Exchange that grades its own callsNext.js 16 · Neon Postgres · AI SDK v7 · Gemini Live · TelegramLive · sign-in ↗Own product
- Community SignalTwo small internal tools that each replaced a weekly manual taskPython · SQLite · Claude / Next.js · monday.com GraphQL · GeminiTwo small internal tools that each replaced a weekly manual taskPython · SQLite · Claude / Next.js · monday.com GraphQL · GeminiInternalClient work
Questions about this service
Do you use n8n and Make, or write code?
- Both, chosen per job. If your team will own and edit the workflow, n8n or Make is the better tool because they can see it. If the logic includes rate limiting, retries, deduplication or anything that needs a test, that part is code, and the workflow tool calls it.
What happens when the source API changes?
- The pipeline stops and posts an alert. It doesn't try to interpret the new shape. Each client keeps the previous scraper as a fallback where one exists. This is the design decision that matters most for anything built on an undocumented source.
Next step
Have a process that runs on someone copying data between two tabs?
Describe it: what it is, who does it, how often, and what goes wrong when it's late. I reply within one working day with a scoping call or a reason it isn't worth automating.