Data Intellect

We started Data Intellect because someone had to say it out loud.

The AI hype cycle has produced more confused executives than working systems. We're two engineers who got tired of watching it from inside the rooms where the bad decisions were getting made. So we left, and started this.

Why we started

By the time 2025 happened, both of us had spent the better part of a decade inside large companies — building software, putting bots into production, sitting in meetings where AI strategy was being decided.

Most of those meetings were broken in the same way. The vendors talking the loudest had shipped the least. The decks looked confident; the production systems behind them, when they existed at all, were brittle and undocumented. Buyers — smart, technical, senior buyers — were being walked into expensive multi-year commitments based on demos that had been hand-tuned for the room.

We watched this happen from both sides. Kinjal had spent eight years inside MNCs building real products, AI-enabled and otherwise, and could tell within the first ten minutes of a vendor pitch which of them had ever actually shipped to production. Disha had spent eight years on the automation side, working with Fortune 200 clients across the world, and had seen the same pattern over and over: a flashy proof-of-concept that quietly fell apart in month two, replaced by a manual process nobody admitted to.

Both of us kept getting asked the same thing by people we trusted, in private: can you just take a look at this and tell me if it's real?

After enough of those conversations, the consultancy started to feel inevitable. We could be the people whose job it is to tell you whether the thing in front of you is real. And then, when it isn't, build you the thing that actually is.

That's what Data Intellect does. We started informally in 2025, took on enough engagements to know it worked, and incorporated formally in 2026.

What we believe about AI right now

A few opinions, stated plainly. You'll see versions of these show up in every engagement we do.

AI is not the answer to every problem, and saying so out loud is part of the job.

A non-trivial share of "AI projects" we get hired to do end with us recommending a database query, a scheduled job, or — occasionally — a better-trained team. When AI is genuinely the right tool, it's powerful. When it isn't, it makes things slower, more expensive, and harder to debug. The difference matters and somebody has to be willing to say which is which.

The interesting AI work right now is at the integration layer, not the model layer.

The frontier-model labs are doing extraordinary work and we use their models heavily. But the leverage for most businesses — yours, probably — is not in training your own model. It's in connecting the right model to the right data, behind the right interface, with the right guardrails. That's mostly engineering, and it's the part most agencies skip because it isn't glamorous.

Smaller businesses deserve the same engineering discipline as enterprises.

There's a tier of AI work in India and globally where MSMEs and startups get sold either off-the-shelf SaaS that doesn't fit their workflow, or hand-rolled prototypes that fall apart in production. We don't think that's a fair trade. The discipline that produces a robust enterprise system — proper evals, observability, version control, runbooks — also produces a robust system for a fifteen-person company. We do both, with the same standards.

Most AI projects fail before any code is written.

They fail in the discovery phase, when nobody asked the boring questions about data, about workflow, about what success will actually look like in month six. We over-invest in that phase on purpose. It's why our pilots ship.

Honesty is cheaper than rework.

We tell clients what's going to be hard, what we're not sure about, and what we'd do differently if we were starting over. It costs us occasional engagements where someone wanted a more confident pitch. It saves everyone money in the long run.

The founders

Two of us. Both engineers. Both have been in production systems long enough to be unromantic about how they actually work.

KN

Kinjal N

Co-founder. Heads engineering and AI development.

Kinjal is a B.Tech-trained full-stack engineer with eight years of experience shipping products inside MNCs across multiple industries. Before Data Intellect, she led builds for products with and without AI components, working across the stack from data layer to UI. Her work has spanned domains including [DOMAIN PLACEHOLDER 1], [DOMAIN PLACEHOLDER 2], and [DOMAIN PLACEHOLDER 3].

At Data Intellect, Kinjal leads the engineering practice. If you've engaged us for a custom AI build or an AI-enabled web product, she's likely the one designing the architecture and reviewing the code that ships.

Writes about

Production AIRAG & retrievalLLM evalsIntegration layer
Kinjal writes about: production AI engineering, RAG and retrieval design, evals for LLM systems, the integration-layer problems that make most AI projects harder than they look.

LinkedIn link pending — INPUT NEEDED (Task 2.4 / Task 3.9 / Task 6.3 all consume this).

DD

Disha D

Co-founder. Heads consulting and automation.

Disha is a B.Tech-trained certified RPA engineer and technical consultant with eight years of experience deploying automation systems for Fortune 200 clients across multiple geographies. Before Data Intellect, she built and shipped automation programs at scale — including [SCALE/EXAMPLE PLACEHOLDER] — working across [GEOGRAPHIES PLACEHOLDER, e.g., "the US, UK, and Southeast Asia"].

At Data Intellect, Disha leads the consulting and automation practice. She's typically your first conversation with us — discovery, scoping, and the upfront work where we figure out whether we're actually a fit for what you need.

Writes about

Intelligent automationRPA + LLM agentsEnterprise automationDiscovery & scoping
Disha writes about: intelligent automation, the boundary between traditional RPA and LLM-based agents, what enterprise automation buyers wish their vendors would tell them.

LinkedIn link pending — INPUT NEEDED (Task 2.4 / Task 3.9 / Task 6.3 all consume this).

How we actually run things

Eleven people including the founders, as of early 2026.

Engineering is run out of Gujarat, India. That's where most of the build work happens, where code reviews live, and where the team is in a room together when we need to be. Sales, scoping, and client-facing conversations are run out of Dubai — closer to the GCC and a more practical timezone for clients in Europe, the UK, and the eastern US.

We work async by default. We're on Slack and email during your working hours, whichever continent you're on. Calls happen when calls are useful — and only when calls are useful. We've heard enough complaints about consultancies that schedule a meeting to schedule the next meeting; we don't do that.

Engagements range from two-week audits to multi-quarter retainers. We don't have a minimum project size, but we do have a minimum project quality — meaning if we don't think we can deliver something we'd be proud of, we'll say so and not take the work.

How we work with clients

Some specifics that come up in first calls often enough that we've started writing them down.

Ownership

You own everything.

Code, models, weights, prompts, evals, infrastructure-as-code, the runbook. We don't keep clients locked in by holding the artifacts. If you want to take a system in-house six months in, we'll help you do it.

We sign the boring things.

NDAs, MSAs, DPAs, security questionnaires. We've done these often enough that turnaround is fast. We don't try to push our paper if you'd rather use yours.

We invoice in INR or USD.

Whichever makes more sense for your tax position. Most international clients invoice in USD; Indian clients usually invoice in INR. GST and other taxes are added per applicable rules; we'll be clear about all of it before you sign anything.

We're not exclusive to a tool stack.

We work across AWS, Azure, GCP, and on-prem. We use OpenAI, Anthropic, Google, Mistral, and open-weight models depending on what fits. RPA work spans UiPath, Power Automate, n8n, and custom Python. The right tool wins; we don't have a vendor that pays us to recommend it.

Closure

Engagements end cleanly.

When work wraps, you get a handover document, a final eval report, and a list of the things we'd watch in the next six months if we were running the system ourselves. No surprise extension invoices, no auto-renewing retainers without your explicit confirmation.

Handover docs
Eval report

If any of this sounded like you wanted to keep reading.

We don't do high-pressure sales calls. The first email is informal, often answered by one of the founders directly, and you'll know within twenty minutes of conversation whether we're a fit.