:

CommVersion runs managed live-chat and AI-assisted conversation products for websites across hundreds of client accounts. Every chat, greeting, lead and page view lands in our warehouse, and the data team turns that into the client dashboards, executive reporting and revenue models the business runs on. We are looking for an Analytics Engineer to own models in our dbt codebase end to end — from raw loaded tables through to the marts behind our Metabase and Power BI dashboards. This is a hands-on build role, not a reporting-request queue. You will write SQL and Python that runs in production every hour. This suits someone roughly 1–3 years in who already writes real dbt models and solid SQL, and wants to go deeper on data modelling, orchestration and warehouse performance with a small, senior team around them.

Our Stack

You will work in all of this. We are not asking you to already know all of it — we are asking you to be strong in SQL, dbt and Python, and to pick up the rest fast.

Duties:

Layer

What we use

Warehouse

Amazon Redshift

Ingestion dlt (Python) — our standard for loading across every source

Transformation dbt — a layered codebase of ~110 models: incremental merges, snapshots, dbt-utils, dbt tests

Orchestration

Apache Airflow 3 — layered daily/hourly DAG chains, run/test gates per model, failure alerting

Python

API extractors and integration scripts (GA4, Google Ads, Zoho, OneDrive, chat platforms), plus scripts that call existing AI/ML services for chat scoring, NER and transcript quality

BI

Metabase and Power BI

Engineering

Git + GitHub, PR review, uv, ruff, sqlfluff, pre-commit, dbt-checkpoint, CI on every PR

Deployment

CI/CD pipelines using Docker, Kubernetes, ArgoCD and GitHub Actions

What You’ll Do

- Build and own dbt models across our staging, intermediate, fact/dim and reporting layers — including incremental models with merge strategies, SCD-style history tables and change-log driven refresh logic.
- Model the domain properly — clients, agents, products, chats, leads, traffic and revenue — with dimensional thinking, sane grain and keys that survive contact with real data.
- Write dbt tests and data-quality checks (generic, singular and custom) so that a bad upstream load fails loudly instead of quietly reaching a client dashboard.
- Build ingestion with dlt — new sources,



incremental/persistent vs non-persistent load patterns, schema evolution, and migrating our remaining legacy connectors over.
- Write Python where the tooling stops — pull from third-party APIs that dlt doesn't cover, handle pagination, rate limits and retries, land the data in Redshift, and call our existing AI services for enrichment.
- Orchestrate in Airflow 3 — author and maintain DAGs, wire dependencies between load, transform and report stages, and keep run times and retries under control.
- Build dashboards in Metabase and Power BI on top of your own models, and work directly with commercial and operations stakeholders on what they actually need.
- Care about performance — distribution and sort keys, join strategy, incremental windows, and why the daily run got 40 minutes slower this week.
- Work like an engineer — branches, PRs, code review, pre-commit checks and documented models. Nothing reaches production straight from a SQL client.

What We are Looking for

Three tiers, and we mean them literally. The first is the bar. The second makes you a stronger candidate, but you will not be rejected for missing it. The third we teach you here — please don't screen yourself out over it.

Must have

- 1+ year in an analytics engineering, data engineering or BI engineering role (internships and academic projects don't count towards this).
- Strong SQL — window functions, CTEs, multi-table joins, aggregation at the right grain, and the ability to read someone else's 200-line query and explain what it does.
- Working Python — requests/pandas, clean functions, error handling and logging. Enough to own a script that runs unattended every morning. • Data modelling fundamentals: facts vs dimensions, grain, and keys that hold up once the data gets messy.
- A branch-and-pull-request workflow — your work reaches production through code review, not a SQL client.
- Clear written and spoken English — you will talk directly to non-technical stakeholders. Strongly preferred These make you a stronger candidate.



None of them is a rejection on its own.
- Hands-on production dbt — ref/source, materializations, incremental models, tests, macros and Jinja.
- Any orchestrator: Airflow, Dagster, Prefect, Mage.
- Any EL tooling: dlt, Fivetran, Airbyte, Meltano.
- Dashboards built in any BI tool (Metabase, Power BI, Tableau, Looker, Superset) with real stakeholders on the other side.
- Redshift, or another columnar MPP warehouse — Snowflake, BigQuery, Databricks.
- GA4 / BigQuery export, Google Ads or other marketing data sources.
- Working with LLM or ML service outputs as a data source — scoring, classification, NER. We'll teach you here Nobody arrives knowing these. They are learned by owning a pipeline in production — which is exactly what this role gives you.
- Slowly changing dimensions and history modelling — keeping last quarter's numbers correct when today's rates change.
- Idempotency, backfills and late-arriving data — why a job that runs twice must not double your revenue.
- Redshift performance: distribution and sort keys, VACUUM/ANALYZE, the SUPER type.
- dlt, Airflow 3, dbt-checkpoint, sqlfluff and CI for dbt.
- Docker, Kubernetes, ArgoCD, GitHub Actions and our CI/CD pipeline.
- AWS: S3, IAM, ECS/EC2, Secrets Manager.
- Our domain — what a chat, a greeting conversion, a lead and a client product actually mean. That takes everyone a few weeks.

Why This role

- Small data team, real ownership — your models are the ones the business reports on, and you'll see them used the week you ship them.
- A genuinely modern stack, kept current: dbt, Airflow 3, dlt, CI on every PR, and an active migration off legacy tooling that you'll help drive.
- Exposure across the full path — ingestion, modelling, orchestration, AI-enriched data and BI — rather than one narrow slice of it.
- Senior engineers who review your code properly and expect you to grow rapid.
- A proper deployment path — CI/CD with Docker, Kubernetes, ArgoCD and GitHub Actions.

You'll work alongside it and pick it up here; we don't expect you to arrive knowing it.

To apply, send your CV along with a short note on one dbt model or data pipeline you built, what made it hard, and what you would do differently now. Links to GitHub, a dbt project or a dashboard you're proud of are welcome and count for more than a longer CV.

📌 Analytics Engineer (India)
🏢 uKnowva
📍 India