The Indian direct-to-consumer commerce ecosystem is entering a new phase, one where speed alone is no longer enough. In categories such as fresh meat and seafood, enterprises are increasingly being forced to rethink how technology, supply chains, fulfilment, and customer experience work together in real time.
For Licious, that challenge is even more complex. Unlike conventional ecommerce models, perishables operate under strict constraints where freshness, timing, inventory movement, and fulfilment efficiency directly impact customer trust and profitability.
In an interaction with Express Computer, Gaurav Mathur, CTO, Licious speaks about how the company is building an AI-first architecture, why real-time event-driven systems are becoming essential for modern commerce, and how autonomous supply chains could define the next phase of digital transformation in D2C businesses.
Building an AI-first architecture where freshness is a “hard constraint”
According to Mathur, the biggest challenge in the perishables business is that freshness cannot be treated as a soft operational metric. “In a category like fresh meat and seafood, freshness is not a metric; it’s a hard constraint. Every minute of delay directly impacts product quality, customer trust, and unit economics.”
Because of this, traditional batch-driven systems are no longer sufficient. Instead, enterprises must move towards continuous, signal-driven architectures capable of making real-time decisions. “Building an AI-first architecture in this environment requires a fundamental shift: from periodic, batch-driven systems to continuous, signal-driven decisioning,” he says.
At Licious, the technology backbone has been designed to process millions of live operational signals simultaneously across demand, inventory, logistics, and fulfillment layers. These signals feed decision layers that operate in near real time thus enabling sub-second responses across critical workflows like allocation, fulfilment, and customer experience.
For Mathur, AI cannot simply sit as an external intelligence layer. Instead, it must become part of the operational core itself. “AI is not a layer on top of the system; it is embedded into the core operating loop. But AI only creates value when it is tightly coupled with execution.”
Real-time commerce requires event-driven architectures
Mathur explains that one of the most significant architectural shifts at Licious has been the move towards an event-driven architecture, where every meaningful system action emits a signal.
Whether it is a user browsing products, an order being placed, or inventory movement across nodes, every event becomes part of a larger streaming ecosystem. Procurement, processing, pricing, fulfilment, and customer-facing systems all now operate on a common real-time operational view.
This architectural transformation has significantly reduced decision latency across the organisation. More importantly, it has helped eliminate fragmented decision-making caused by stale or delayed operational data.
Machine learning is helping Licious move closer to just-in-time operations
Demand forecasting remains one of the most difficult challenges in perishable commerce. Unlike traditional retail, demand patterns can shift suddenly based on weather, local events, behavioural changes, or supply disruptions.
Mathur believes that forecasting is less about perfect prediction and more about continuous adaptability.
Licious uses machine learning models across highly granular operational layers, including pin codes, SKUs, and time bands.
But static forecasting models alone are insufficient. Real-time demand sensing has become equally important. “External signals like weather, local events, and behavioural shifts can quickly invalidate static forecasts. Our systems are designed to continuously recalibrate using live data,” he points out.
The forecasts themselves are deeply integrated into operational decision systems rather than functioning independently.
This helps the company move closer towards just-in-time operations while reducing wastage and stockout risks simultaneously.
Transition from monoliths to microservices
Like many fast-growing digital businesses, Licious is also transitioning from monolithic systems to distributed microservices architectures.
But according to Mathur, this journey involves far more than just technology modernisation. “The move from monoliths to microservices is not just a technical transition; it’s an organisational one.”
While microservices improve flexibility and scalability, they also introduce new forms of operational complexity.
Instead of attempting large-scale rewrites, Licious adopted a phased modernisation approach. “We approached this pragmatically, starting with high-impact domains and progressively decoupling systems.”
At the same time, the company invested heavily in observability frameworks, centralised logging, distributed tracing, and real-time monitoring.
The ownership models also became critical in maintaining distributed system reliability at scale. “Today, we operate with clear service ownership models, where teams own services end-to-end,” he says.
One of the most important lessons from this transition, he says, is understanding the balance between scalability and operational coordination.
Personalisation extending beyond recommendations
For Mathur, modern personalisation is no longer restricted to recommendation engines. “Personalisation today extends far beyond recommendations. At Licious, it spans the entire customer journey — from discovery and pricing to fulfilment and post-purchase engagement.”
But unlike many digital businesses, Licious cannot optimise personalisation independently from supply chain realities. This means customer experiences are dynamically influenced by operational feasibility in real time.
“What a customer sees in the app is dynamically influenced not just by preferences, but also by real-time fulfilment feasibility,” he affirms.
AI-driven decisioning systems now continuously balance customer relevance with operational efficiency. “We’ve built systems where personalisation decisions are made in real time. This has enabled measurable improvements in key metrics like conversion and average order value while maintaining fulfillment reliability.”
Cloud optimisation is now as important as scalability
As AI workloads and real-time systems become increasingly compute-intensive, cloud strategy has become central to operational efficiency. Licious follows an elastic scaling strategy supported by continuous optimisation. “Our cloud strategy is built around elastic scaling combined with continuous optimisation,” he says.
The company uses a mix of spot and on-demand infrastructure to balance cost efficiency with scalability requirements. “We’ve reduced infrastructure costs by around 40 per cent through workload right-sizing, architectural optimisations, and eliminating inefficiencies. At the same time, we improved latency across critical APIs by 2–3x.”
Reliability, however, continues to remain foundational. “We design systems with redundancy, failover mechanisms, and strict latency guardrails.”
Autonomous supply chains could define the next phase of D2C commerce
Looking ahead, Mathur believes the future of commerce will not be defined simply by better consumer-facing applications. “The next phase of D2C will not be defined by better apps. It will be defined by autonomous and self-correcting supply chains.”
Enterprises are steadily moving towards AI-led operational systems capable of independently handling demand sensing, inventory allocation, and delivery orchestration.
In practical terms, this means systems capable of dynamically adjusting production, rerouting inventory, and anticipating demand before orders are even placed.
For Mathur, the real transformation lies in creating closed-loop systems where data, decisions, and execution continuously learn from each other.
Ultimately, he believes AI is evolving beyond being a feature into becoming the operational backbone of modern commerce businesses. He concludes, “At Licious, we see AI not as a feature, but as the operating system of the business. In categories where freshness, speed, and reliability define customer trust, this is not just a competitive advantage; it is a necessity.”