How Blue Star's CIO is shaping an AI-ready data estate

Time of India

2 days ago

In an industrial world driven by precision, performance, and unpredictability, data has emerged as the new compass. Yet, most enterprises are still wrestling with fragmented, siloed data that refuses to speak the same language. Udit Pahwa, the CIO of Blue Star, has been on a quiet mission to change that.Across his three-year journey, Udit has been building an 'intelligent data estate' not as a buzzword but as a practical, measurable foundation to deliver AI-led business outcomes. In an exclusive DeepTalks session, he took us behind the curtain, into the decisions, mistakes, frameworks, and breakthroughs that have shaped Blue Star's transformation.
The data deluge that reaches the boardroom
'Traditionally, data was structured and lived comfortably within ERPs and CRMs,' Pahwa begins. 'But today, the universe of unstructured and real-time data, from social media, call centers, factory sensors, is exploding in volume and richness.'
The challenge, Pahwa explains, is no longer about having data, but connecting it. Structured data is predictable. Unstructured data is ambiguous but full of context. Real-time data never waits. The power lies in unifying them, if only organizations can figure out how.
'This isn't just a CIO's headache anymore,' Pahwa adds. 'It's a boardroom discussion. Boards want agility, frugality, and data-driven decisions. They expect new revenue streams from data-as-a-service models. That's why this matters.'
Building the intelligent data estate
So, what makes a data estate 'intelligent'? Pahwa breaks it down into a triangle: Architecture, governance, and outcomes.
Architecture is the plumbing, the cloud-native infrastructure, the streaming platforms, the integration logic, and the decisions around centralized or decentralized models.
Governance is the invisible shield, defining data quality, security, privacy, and ownership. 'If you don't get governance right,' Pahwa says, 'you'll end up with garbage in, garbage out.'
And then come the outcomes. 'Ultimately, why are we doing all this? We must tie every data initiative to a measurable business benefit, higher NPS, faster launches, more uptime, more revenue.'
Blue star's journey
Blue Star's journey didn't begin with flashy tools. It began with purpose.
'Three years ago, we started with structured data,' Pahwa shares. 'ERP and CRM were reliable sources. We focused on sales and service. Once that base was strong, we began ingesting unstructured data, social media, voice-to-text from our call centers, for sentiment analysis.'
Today, they're working on real-time ingestion from the shop floor. It's an ongoing evolution, not a sprint.
'We're now capturing sensor data like temperature, vibration, and humidity. Combined with digital maintenance logs, this helps us predict failures before they happen. That's real ROI,' he says.
Lakehouse or mesh?
Centralized lakehouses are powerful, Udit says, they simplify storage, analytics, and governance. But they can also become bottlenecks.
Decentralized data meshes allow domain-specific insights but require stronger cultural shifts and governance.
'So what's the answer? Hybrid,' Pahwa says. 'You take the best of both worlds. Maintain a centralized lakehouse where needed, but give autonomy to business
domains
with proper guardrails.'
At Blue Star, they've chosen centralization, for now, but they're future-ready to adapt as needs evolve.
The vector question
Pahwa cautions against jumping on the vector database bandwagon just because it's hot. 'Ask: Do you need semantic search? Personalization? Anomaly detection? If yes, then run a pilot.'
He explains vector databases in simple terms: 'It's like attaching numbers to your data and storing similar meanings together. You get faster, smarter searches. But only if your use case demands it.'
One powerful example Pahwa cites is using GenAI for shop floor technicians. 'Imagine uploading all user manuals, and the technician just asks a chatbot: 'I'm stuck with this belt issue, what do I do?' That's AI with purpose.'
Real-time or batch? Let business value decide
'Real-time data is expensive,' Pahwa admits. 'It demands more memory, network, compute, and monitoring.'
So how does Blue Star decide?
'We go real-time only where it matters, machine safety, customer experience, dynamic pricing. For strategic planning and long-term insights, batch works just fine.'
The hidden culture shift
Perhaps the most underappreciated part of Blue Star's transformation is the cultural shift.
'Cloud-native architecture changed our operations,' Pahwa says. 'But fostering a data-driven mindset across the company? That's been our biggest pivot.'
Democratizing data ownership and literacy, making sure the business, not just IT, owns data decisions, has created long-term change.
Governance without bureaucracy
Governance is often seen as a hurdle. Pahwa's fix? Governance-as-code.
'Treat policies like software. Automate schema enforcement and validation. Use data ops pipelines. It shouldn't feel like red tape, it should just work,' he says.
DPDP and consent management
With India's DPDP Act, privacy isn't optional anymore. Blue Star is preparing with a centralized consent management platform.
'You need revocable consent, transparency, and embedded privacy by design. This isn't just compliance, it's trust,' he says.
Vendor selection
Pahwa emphasizes long-term partnerships over transactional deals. His layered approach:
Hyperscalers for infraData platform specialistsML and vector database vendors
Evaluation goes beyond price: support models, IPRs, privacy, exit clauses, all count. 'The best vendor is one who aligns with our business outcome,' he says.
FinOps and the cloud cost dilemma
'Cloud costs can spiral fast,' Pahwa warns. His three-pillar mantra:
Visibility: FinOps tools to track every GB, every query.Optimization: Right-size resources. Use serverless where possible.Accountability: Make cost consciousness a cultural trait.
The future of data estates
In the next 3 - 5 years, Pahwa sees:
Self-optimizing pipelines: Data lakes that fix themselves.Edge + GenAI: Real-time intelligence at the source.AI literacy everywhere: Prompt engineering will be as common as Excel.Federated enablement: Domains will own pipelines, but guardrails will remain centralized.
His message to fellow CIOs and CISOs?
'Start with enablement, not control. Build self-service platforms. Invest in automation. And above all, focus on business value,' he says.
As the conversation ends, Pahwa reflects on how far organizations have come.
'Data is no longer a back-office topic,' he concludes. 'It influences everything, from strategy to execution. And it's not just a corporate issue anymore. In our personal lives too, we're learning to trust data for decisions.'