Zohaib Aamir

About

Eight years building data platforms, and the AI systems that sit on them.

I am a data and AI engineer. The platforms came first: warehouses, lakehouses, pipelines and the BI on top of them, on Azure and AWS, for teams up to thirty engineers. The agents came after, and they only worked because of what was underneath.

In the last two years I have led the AI and data track on a multi-agent system for 5,000 dealerships, been tech lead on a compliance platform in a safety-critical domain, led a three-person team that shipped a retrieval assistant now in production, and designed the quality gateway that twenty-six markets deliver into. Each one is written up here, including the parts that did not go to plan.

Each engagement ends with a handover: runbooks, documentation, and the client's own engineers running the system.

Role
Data and AI Engineer
Based
Remote
Cloud
Azure, AWS, GCP
Largest team led
30 engineers
Industries
Automotive, Rail, Pharma, Life sciences, Healthcare, Finance, Manufacturing, Oil and gas, Aerospace, Gaming

How I work

What production means, in practice.

Six things every system here has to have before anyone signs it off. Each one links to where it is implemented.

Permissions

Scope comes from the authenticated session and is applied inside the retrieval query or the SQL, before any result exists.

Evaluation

Outputs are checked against known answers before a release and after every prompt, model or schema change.

Monitoring

Every run is traced. A failed sync or a failed check raises an alert instead of failing silently.

Declining to answer

When the system cannot support an answer it says so and routes to a person, rather than producing confident text from thin sources.

Human review

A qualified person makes every decision that reaches a customer, and that decision is recorded as a label the system learns from.

Audit trail

Every answer, action, skipped review and version is written with the user and the time, and is immutable once signed.

Experience

Engagements.

Role, context, what I contributed, and what changed.

  1. Automotive

    Started early 2025. Live in production.

    Dealer service agents

    A European automotive distribution group

    Led the AI and data engineering: the multi-agent architecture, the retrieval and text to SQL pipelines, the embedding fan-out, and the governance layer.

    Outcome

    Questions a week across 5,000 dealerships, each scoped to the dealer asking

    Azure OpenAI, LangGraph, LangSmith, pgvector, Azure Functions, Azure Container Apps, Apache Superset

    Case study

  2. Rail

    First release two months after kickoff, then a release every six weeks. Final signoff in January 2026.

    Requirement compliance platform

    An industrial manufacturer in the EU building safety-critical systems

    Tech lead. I owned the overall architecture design and contributed the AI engineering: the document pipeline, retrieval design, the supervisor and worker structure, and the evidence model.

    Outcome

    To a first full compliance analysis, down from 30 to 50 days

    Azure AI Document Intelligence, Azure OpenAI, pgvector on Azure Database for PostgreSQL, LangChain, FastAPI

    Case study

  3. Pharma

    Rolled out market by market across 26.

    Data quality portal and analytics platform

    A pharma distribution business operating across Europe

    Designed the data quality gateway and the validation model, and led the data engineering and the Fabric platform build.

    Outcome

    Fewer data quality incidents across 26 regulated markets

    Azure Data Lake Storage, SQL database, Microsoft Fabric (data engineering, semantic models), Power BI, data quality portal

    Case study

  4. Manufacturing

    Discovery from December 2024. Development from June 2025. Releases in July, October, and November 2025.

    Formulation and offer assistant

    A European food ingredients manufacturer

    Lead engineer. I signed off the architecture and led development day to day.

    Outcome

    First reply to a customer, down from at least two weeks

    Azure OpenAI, Azure AI Foundry, PostgreSQL with pgvector, FastAPI, React, Azure Container Apps

    Case study

  5. Finance

    2025.

    Peer and transaction recommendation agents

    A financial software company in the EU

    Built the recommendation pipeline and the MCP server.

    Outcome

    To a ranked peer shortlist, previously assembled by hand across three databases

    Python, PyTorch, Snowflake, AWS Bedrock, MCP, GraphQL

    Case study

Stack

What I build with.

Grouped by job. Most of it is in one of the case studies; the rest is from earlier platform work.

Data

  • Python
  • SQL
  • PySpark
  • dbt
  • Kafka
  • PostgreSQL

Platforms

  • Databricks
  • Snowflake
  • Microsoft Fabric
  • Azure Synapse
  • Azure Data Factory
  • BigQuery
  • Redshift

AI and agents

  • LangGraph
  • LangChain
  • LangSmith
  • Azure OpenAI
  • Azure AI Foundry
  • AWS Bedrock
  • pydantic-ai
  • MCP
  • PyTorch

Retrieval

  • pgvector
  • Azure AI Search
  • Azure AI Document Intelligence

Cloud and serving

  • Azure
  • AWS
  • GCP
  • FastAPI
  • Azure Functions
  • Azure Container Apps
  • React

Analytics

  • Power BI
  • Apache Superset
  • Looker

Contact

Discuss a project.

Tell me what you are building and what is in the way.

Elsewhere

Opens in your mail client, addressed to zohaib@zohaibaamir.com.