Company background 

‍

Tata 1mg serves millions of customers with affordable medications and operates at national scale within a $200B conglomerate. Prescriptions are central to their commerce flow. Historically, the digital prescription step required a pharmacist to manually review, tag, and approve orders, creating a throughput ceiling just as demand surged. The company’s growth mandate was to scale safely, lower unit costs, and comply with strict data privacy which made automation of the prescription workflow a board-level priority.

‍

The Problem

‍

Manual processing created both a throughput cap and revenue leakage:

‍

• High drop-offs: Orders stalled for hours while waiting for manual tagging, causing customers to abandon carts.
• Throughput limits: Manual tagging capped output at ~300k prescriptions/month vs. ~800k+ demand.
• Unit economics: With an AOV ~ $12, manual digitization consumed 1–2% of AOV, yielding millions in avoidable costs.
• Infra constraints: Infra bottlenecks: Slow, DevOps-heavy deployments; no autoscaling; and no reliable setup for OCR + vision-language experimentation.

‍

Bottom line: The team needed fast, accurate, and compliant prescription understanding in-house with the headroom to keep up with growth.

‍

Solution

‍

Tata 1mg adopted Simplismart’s MLOps platform to own the end-to-end lifecycle across experiment → fine-tune → deploy → observe without waiting on DevOps tickets. The engagement focused on three pillars:

‍

1) Inference & serving to make every millisecond count

‍

• 5× faster inference: Simplismart’s engine sped up VLM processing and stabilized latency for consistent SLAs.
• Zero queueing: Containerized deployments + autoscaling let capacity expand with upload spikes, removing the long stalls.

‍

2) Training & experimentation to ship better models, faster

‍

• Moved to open-source models: Replaced paid APIs with fine-tuned VLMs that handled handwriting and noisy scans better.
• Much faster training: Distributed training cut cycles from days to hours, enabling rapid iteration.
• Faster deployments: Data science deployed directly, reducing time-to-production from months to ~2 weeks.

‍

3) Observability, privacy & compliance to operate with confidence

‍

• Moved to open-source models: Replaced paid APIs with fine-tuned VLMs that handled handwriting and noisy scans better.
• Much faster training: Distributed training cut cycles from days to hours, enabling rapid iteration.
• Faster deployments: Data science deployed directly, reducing time-to-production from months to ~2 weeks.

‍

Results

‍

• Accuracy: Prescription parsing jumped from ~80% to ~95%, with pharmacists only handling edge cases.
• Velocity: Model launches dropped from months to ~2 weeks; training shrank from 3 days to ~1.5 hours.
• Capacity & CX: Manual 2-hour stalls became near-real-time tagging, reducing drop-offs and recovering revenue.
• Cost & control: High-cost APIs and manual tagging were eliminated, improving unit economics and keeping data in-house.
• Reliability: Observability, SLAs, and autoscaling kept latency stable and absorbed traffic spikes without queues.

‍

With Simplismart, Tata 1mg’s prescription flow shifted from a manual bottleneck to an instant, automated pipeline. Parsing is now real-time, accuracy is enterprise-grade, and new model upgrades ship in days not months.

‍

Engineering teams experiment freely, autoscaling absorbs demand spikes, and pharmacists intervene only for true edge cases. Simplismart helped Tata 1mg unlock predictable SLAs, lower costs, and deliver seamless medication ordering at national scale.