Taking two new engagements for Q4 2026

Get the reporting right. Then put AI on top.

We build the analytics estate and the AI that stands on it: pipelines and dbt models, dashboards leadership actually opens, product analytics that settle arguments, and agents that watch the business without making things up. For anyone who eventually gets asked how the number was produced.

Executive dashboard live in 21 days Evals on every prompt change 30 days notice, no exit fee
01 — SOURCE Systems, events, third parties 02 — CONTEXT dbt metrics, contracts, docs 03 — AI SYSTEM Retrieval, tools, agent 04 — GUARDRAILS Evals, grounding, tracing 05 — DECISION Logged, explained, monitored
MicrosoftSonyWells FargoFidelity InvestmentsHudson's BaySobeysRoyal Bank of CanadaBank of Tokyo-MitsubishiMicrosoftSonyWells FargoFidelity InvestmentsHudson's BaySobeysRoyal Bank of CanadaBank of Tokyo-Mitsubishi

Where our leadership has delivered · 2011–2026

The problem

Three ways an AI project stalls

Almost never at the model. Nearly always in the six inches between the model and the data it was pointed at.

Stall one

The demo worked

A model wired to a copy of the data answered ten questions beautifully in a meeting. Nobody could say what it would do on the eleventh, so it never left the meeting.

Stall two

The answers are plausible

It writes SQL against a schema it has to guess at. Sometimes it is right. You find out which times only when somebody in finance happens to notice.

Stall three

Nobody will sign it off

Risk asks how you would explain a declined applicant eighteen months from now. There is no trace, no eval and no version history. The project stops there, correctly.

What actually fixes it

The model is not the hard part. The context is.

An assistant pointed at a normalised warehouse has to reconstruct years of business logic from table names, and it will do it confidently. Give it a governed metric layer instead of raw schemas, tools instead of free-form SQL, and an eval suite that runs on every prompt and data change — and the same model becomes something you would put in front of a regulator.

The method
The method

One working thread first

A thread is an engineering term: the thinnest end-to-end slice of a system that proves the whole path works. It is also how we sell, because it is the only honest way to find out whether this holds up in your business.

One decision, wired from source data through context, model and guardrail into production. Thin, complete, provably working. Then we widen it.

If the thread holds, we widen it. If it does not, you have spent six weeks rather than six months — and you have learned something true about your data.

01

Fit call

30 minutes

You describe the decision you want to improve. We tell you whether AI is the right tool for it, and say so plainly when it is not. No deck.

02

Readiness Check

2 weeks

A fixed-fee diagnostic. Where AI would pay, whether your data can carry it, what any existing AI would fail on, and a prioritised 90-day plan you own outright.

03

The first thread

3 to 6 weeks

One decision, wired end to end: source data, context layer, model, guardrail, production, monitoring. Deliberately narrow. It either works or it tells us something true.

04

Widen it

Ongoing

Once the thread holds, the pod broadens it on 21-day cycles. Every cycle ends with something in production and a written readout. Thirty days notice, no exit fee.

The pods

Two families. Most people start with the first.

Analytics pods build the estate — pipelines, models, dashboards, product analytics. AI pods build what stands on it. Buying the second without the first is the most common and most expensive mistake in this market.

Start here

AI & Data Readiness Check

An honest answer to two questions: where AI would actually pay in your business, and whether your data can carry it yet.

€7,500 fixed fee · 2 weeks
If you go on to run a pod with us within 60 days, the fee comes off your first invoice. The plan is yours either way.

Two weeks, and you own the output

  • Interviews with the people who use, and mistrust, the numbers
  • A review of your pipelines, dbt project, semantic models and reporting
  • Any AI already in the path, and the specific things it would fail on
  • Your EU AI Act exposure, if you have any, in plain language
  • A scored assessment across the five capability areas
  • A prioritised 90-day plan, with the first working thread specified
What changed

Every analytics role has been rewritten in eighteen months

Not replaced — rewritten. We staff and build for what these jobs are now, rather than what they were on the last org chart.

Analytics engineers

Have quietly become the most important people in the building. The dbt project is no longer just what feeds a dashboard — it is the context an AI system reads. A metric defined once and tested is a metric a model cannot get wrong.

BI developers

Spend less time building pages and more time preparing semantic models for Copilot: synonyms, hierarchies, descriptions, verified answers. Most Copilot disappointment is a model preparation problem wearing a Copilot costume.

Analysts

Draft with a model and validate by hand. The work moved up a level: less writing the query, more deciding what the question means and whether the answer can be trusted.

Data engineers

Scaffold pipelines and generate quality tests with assistance, then spend the time they got back on contracts, lineage and the freshness guarantees an AI system quietly depends on.

AI engineers

A role that barely existed three years ago and now sits in every serious data team. Retrieval, tool use, agent boundaries, evals — the systems work around a model somebody else trained.

Model risk and governance

Moved from an annual validation document to a live discipline: inventories, model cards, eval evidence and, since August 2026, Annex III obligations for anyone scoring creditworthiness.

Track record

Built by someone who has shipped this in production

ProAnalytiqs is led by Madhukar Reddy V, who has built analytics and AI programmes for Microsoft, Sony, Wells Fargo, Fidelity Investments and Hudson’s Bay, and now runs the analytics and AI stack for a multi-market European fintech. The figures below are his, not a client average.

We do not publish client outcome numbers. Ask on a call and we will walk you through the work in detail, under NDA where it needs to be.

Read the full profile
0
Years building data platforms
Banking, fintech, retail and big tech
0
European markets
Reported on daily, one governed model
0
Largest team led
Analysts, engineers and scientists
0
Regulated industries
Retail banking, consumer credit, wealth, capital markets
Madhukar Reddy V, founder of ProAnalytiqs
Bengaluru, India IST (UTC+5:30) · four hours of daily overlap with CET
Leadership

Madhukar Reddy V

Founder & Principal AI and Analytics Lead · CFA Level III Candidate

“Anyone can get a model to answer a question. The hard part is being able to say, eighteen months later, exactly why it gave that answer — and most teams find that out far too late.”

Thirteen years building analytics, BI and data engineering for Microsoft, Sony, Wells Fargo, Fidelity Investments, Hudson's Bay and Royal Bank of Canada. Now leads the analytics and AI stack for a multi-market European fintech, with generative AI running in production against it.

  • Led analytics teams for Microsoft, Sony, Wells Fargo and Fidelity Investments
  • Analytics strategy for Hudson’s Bay, Sobeys and Royal Bank of Canada
  • Generative AI, Databricks, Snowflake and Power BI running in production today
  • Direct experience of regulator, audit and funder scrutiny — the reason guardrails come first here
Questions

The things people ask before signing

Are you just putting a chat box on top of our database?

No, and that is the thing we are usually hired to undo. Pointing a language model at a normalised warehouse produces confident, wrong answers, because the model has to guess your business logic from table names. The work is in the layer underneath: metrics defined once in dbt, documented in a way a model can read, and served to the assistant as tools rather than tables. The chat box is the last two percent.

We do not have dbt or a semantic layer yet. Are we too early?

You are not too early to talk, but you probably are too early to buy an AI assistant. That is what the Foundation and Context pods exist for, and it is what the Readiness Check is honest about. Building AI on top of pipelines nobody trusts just moves the mistrust somewhere more expensive.

How do you stop it hallucinating?

Three things, in order. Constrain what it can reach — it queries defined metrics through tools, not arbitrary SQL. Check the answer against the retrieved context before it is shown, so an unsupported claim fails rather than ships. And run an eval suite on every prompt, model or data change, with a golden set of real questions. Hallucination is not a prompt problem, it is a systems problem.

Does our data end up training somebody else’s model?

Not with the setups we build. We default to enterprise endpoints with no training on your data, run inside your own cloud tenancy where the workload allows, and pseudonymise or exclude personal data before anything leaves your environment. If a use case genuinely cannot be built that way, we say so before you sign, not afterwards.

Are we in scope for the EU AI Act?

Anything that materially affects a person — credit, employment, education, insurance pricing, access to essential services — is worth checking, because Annex III high-risk obligations started applying in August 2026. Plenty of AI is not in scope at all. Working out which of your systems are, and what evidence each one needs, is part of the Readiness Check and the whole point of the Guardrail Pod.

All questions answered

Tell us which decision you want to improve.

A 30-minute call, no deck and no pitch. You describe the decision, we tell you whether AI is the right tool for it — and say so plainly when it is not.

Usually a reply the same working day · IST (UTC+5:30)