Announcing Damasqas V1 — The Tool I Wish I Had

I've spent the last three years building data pipelines across every industry you can think of. Today I'm launching the tool I needed at every single one of them.

The same problem, everywhere

I've built data infrastructure at AWS, where I was on the MWAA team — Managed Workflows for Apache Airflow — debugging stuck tasks and pipeline failures for Fortune 500 customers. I shipped trading data systems at Moment, a quantitative hedge fund where a silent gap in market feeds meant trading on stale prices. I wrangled enrichment pipelines at Pocus, a sales intelligence startup that got acquired by Apollo, where vendors would silently degrade and nobody knew until reps complained about bad emails. And I debugged pipelines at an insurance firm where a single missing row could mean a compliance violation.

Different industries. Different stacks. Different scales. But the same problem every single time:

At AWS, I was literally the person F500 companies called when their Airflow DAGs got stuck. I'd dig through CloudWatch logs, cross-reference scheduler heartbeats, check worker resource limits — the same investigation loop, hundreds of times, for customers who were paying millions for managed infrastructure and still couldn't figure out why their pipeline stopped. At Moment, a silent data gap in market feeds meant the desk was trading on stale prices for hours before anyone noticed. At Pocus, enrichment vendors would silently degrade — accuracy would drop from 97% to 68% and nobody would know until a sales rep complained that half their leads had wrong emails.

The pattern is always the same. The data breaks. The alert is either missing, too noisy, or too vague to act on. Someone spends hours digging through logs, cross-referencing deploys, reading workflow histories, and tracing rows across tables — just to find the one line of code that changed.

Why existing tools don't solve this

I tried everything. Observability platforms give you dashboards and traces, but they don't understand data. They know your service is returning 500s, but they don't know that the 500 was caused by a column rename in a migration that went out 12 minutes before the pipeline ran.

Generic AI coding agents are the hot new thing. Give an LLM a sandbox, let it write code. Cool for building a todo app. Useless when you need to trace a single row across four tables, understand why a Temporal workflow retried three times before failing, cross-reference that failure with a GitHub PR that renamed a column, and suggest a fix that accounts for two other workflows that reference the same column.

That's not a coding problem. That's a data engineering problem. It requires domain knowledge that no generic tool has.

So I built Damasqas

Damasqas is an AI data engineer that lives in your Slack workspace. It connects to your databases, your workflow engine, your deployment platform, and your code — and it actually understands the relationships between them.

Ask it a question in plain English:

It didn't just tell me what happened. It traced the root cause across Temporal, Railway logs, and the Clearbit API. It wrote a fix. It pushed a PR. It spun up a sandbox and ran tests. And it asked me to approve — all in Slack, all in under a minute.

That's what I mean by data engineering intelligence. Not "I can write code in a container." I can understand your pipeline DAGs, your data contracts, your vendor SLAs, and what "correct" actually means for your specific domain.

What V1 can do

Today's launch includes everything I wished I had at every company I've worked at:

Ask questions in plain English. Query your database, trace records across tables, inspect workflow histories, check deployment status, browse your code — all from Slack. Damasqas writes the SQL, runs the queries, and explains the results.

Full data lineage tracing. "Where did this row come from?" — get the complete path across every API call, transformation, staging table, and production write. End to end.

Autonomous failure resolution. When a pipeline fails, Damasqas doesn't just alert you. It identifies the root cause, writes a fix, pushes it to GitHub, spins up a dev box to test it, and opens a PR for your approval.

Domain-aware monitoring. Set up monitors in plain English: "alert me if our order count drops below 100 in any hour." Damasqas understands that 847 missing stock rows is critical, that enrichment accuracy below 94% means your sales team has bad data, and that a stale payments table during business hours is a problem but at 2am on a Sunday it's fine.

Rule-based automation. "If revenue-sync fails 3 times in a row, alert #sales-ops." "Every Monday at 9am, post a pipeline health summary." Rules that modify state — like pausing a schedule — require explicit approval before Damasqas acts.

Sandbox PR verification. Every pull request gets tested in an isolated environment before it touches production. Damasqas detects your stack, builds the project, runs your tests, and reports results to the GitHub PR check and Slack.

Intelligent model routing. Not every question needs Opus 4.6. A schema lookup gets routed to a fast, cheap model. A cross-pipeline root cause analysis gets the full reasoning power. You pay for what each task actually needs.

Why this isn't just another wrapper

I've seen the wave of AI coding agents — Vallo, Replicas, and dozens of others. They're useful tools. But they're infrastructure wrappers. They give an LLM a container and let it write code. The value is the sandbox, not the intelligence.

Damasqas is different because the value is the intelligence. It's not a general-purpose coding agent that happens to know some SQL. It's a specialist that understands:

• Pipeline DAGs and workflow orchestration semantics
• Data contracts and what "correct" means for your specific domain
• The relationship between a GitHub PR, a Railway deploy, a Temporal workflow failure, and a Supabase schema change
• That "enrichment accuracy dropped 3%" is a different severity depending on whether you're a sales team or a trading desk

If Anthropic ships a better coding sandbox tomorrow, a wrapper dies. If Anthropic ships a better model tomorrow, Damasqas gets smarter — because the domain knowledge, the integrations, the monitoring rules, and the data context are ours.

What's next

V1 ships with connectors for Supabase, GitHub, Temporal, Railway, Slack, and PostgreSQL. We're already building support for:

• Pipeline orchestration: Apache Airflow, Apache Kafka, Apache Spark
• Databases: Snowflake, RisingWave, StarRocks
• Observability: Grafana, Loki, Prometheus
• Data transformation: dbt

Every new integration is an MCP server that plugs in without changing the bot. The architecture is built to grow.

Try it

Damasqas is free to start — connect one project, ask unlimited questions, get zero-config baseline alerts. If you're an engineering lead or founder who wants the full picture, book a demo and I'll show you what it looks like on your actual stack.

I built this because every data team I've ever been on needed it. If you've ever been woken up at 3am by a broken pipeline and spent an hour just figuring out what broke — this is for you.

Let's go.

— Shalin