Frequently asked questions

It depends on what we're doing. An audit takes two to three weeks. A pilot or proof-of-concept takes four to eight weeks. A full custom build is usually three to six months, occasionally longer for systems that integrate deeply across an enterprise. We share a written timeline with milestones at the end of the planning phase, not as a vague "weeks to months" range.

No. About a third of our work is with MSMEs and early-stage startups. The reason most AI agencies don't take small projects is that they've structured around large ones; we structured to handle both, and the same engineers do both. If your problem is real and your data is workable, project size isn't the qualifier.

We've delivered work for several. The two founders have sixteen years of combined experience inside MNCs and Fortune 200 environments before starting Data Intellect, so the procurement, security, and change-management parts are familiar territory. Happy to walk through how we've handled past enterprise engagements on a call.

Almost certainly not. "We need to fix our data before we can do AI" is the most common reason AI projects never start. We work with the data you have. Part of the audit is figuring out what's usable now, what's worth fixing, and what's a waste of time to chase.

We integrate with what you have. We've worked across AWS, Azure, GCP, on-prem, and hybrid. We've built on top of Salesforce, ServiceNow, SAP, custom internal systems, and various vintages of homegrown software. We'd rather slot AI cleanly into your existing stack than ask you to rebuild around ours.

Data residency, model choice (cloud-hosted vs. self-hosted), encryption, access controls, audit logging — we cover all of it in the planning phase, before any data leaves your environment. For sensitive workloads we deploy open-weight models inside your infrastructure so your data never leaves your perimeter. We'll be straight about what's required for your compliance regime before you sign anything.

You own the code, the models, and the runbook. If you want us to keep monitoring and improving the system, we'll do that on a monthly retainer. If you want to take it in-house, we'll hand off properly — including pairing with your team for as long as it takes for them to be confident running it.

Yes. We do advisory work on retainer and ad-hoc — covering AI strategy, technical reviews of work other vendors have done, evals of production systems that aren't behaving, and second opinions before you sign with someone else. Founder-level rates apply.

We'll tell you, and we'll point you at what is — even when it's not us. This happens often enough that we've stopped feeling weird about it. A good AI consultancy talks people out of AI roughly as often as it talks them into it.

Engineering team is in Gujarat, India. Sales and client-facing work is run out of Dubai. We work asynchronously with clients across India, the GCC, the UK, and the US. Time zones haven't been a blocker for any current engagement; we've structured the team specifically to make this work.

Yes, freely. We've signed enough NDAs that turnaround is fast — usually same-day if you send us your form. If you'd prefer we send ours, we have a one-page mutual NDA that takes about five minutes to read.

About ₹2,50,000 for an audit-and-roadmap engagement. Below that we can't deliver something we'd be proud of. There's no minimum monthly retainer above the ₹3,50,000 starting rate.

Generally no. Fixed-price builds either get padded with contingency (you pay more for predictability) or fail when scope shifts mid-project (you pay for change orders). T&M with a weekly invoiced cap is honest about how real builds behave. We will fixed-price audits and pilots because those have well-bounded scope.

Yes. For sensitive workloads we deploy open-weight models inside your infrastructure — your data never leaves your network. We've worked inside AWS GovCloud-equivalent setups, air-gapped on-prem environments, and bring-your-own-key cloud configurations.

We work within compliance regimes our clients hold. We don't hold SOC 2, ISO 27001, HIPAA, or PCI-DSS certifications ourselves — claiming any of those when they're not audited is dishonest and legally exposed. If your project requires Data Intellect to be the audited entity, we're probably not the right fit and we'll say so up front.

Yes. We have a template DPA — see /dpa — that we can sign as a Processor under GDPR / DPDP Act 2023 / UAE PDPL terms. Most clients prefer to use their own DPA template; we'll review and sign that too.

Most engagements are in India, the GCC (UAE, KSA), the UK, and the US. We've also worked with clients in Singapore, Australia, and Germany. Async-first working style means continent isn't usually the blocker.

Whichever fits the problem. OpenAI's GPT family (4o, o1, etc.) for general-purpose reasoning. Anthropic's Claude family for long-context and code work. Google's Gemini for multi-modal tasks. Mistral and Llama for self-hosted requirements. For embeddings we'll typically benchmark 3-4 options on your data before committing. We're not exclusive to any vendor; the model that works for your evals wins.

Yes, when fine-tuning is genuinely the right call. Most of the time it isn't — better prompts + retrieval beat fine-tuning for less effort and less drift risk. We'll tell you which side of that line you're on after the audit. When fine-tuning IS right (domain-specific terminology, structured output formatting, latency-critical use cases), we handle dataset prep, fine-tuning runs, and eval framework.

Anywhere from 1 to 4 people. Most engagements run with a lead engineer (one of the founders for the first 4-6 weeks of any new account), a junior engineer for execution, and an integration specialist for environment-specific work. We don't pad teams to inflate billing.