High-volume SMT lines faced reactive planning due to zero real-time visibility. Lack of linkage between child components and final PCBs caused massive traceability gaps during root cause analysis.

To digitize shop-floor inventory and engineer a multi-level traceability architecture linking material flow from individual components to the final unit.

Built a QR-based component-to-product linkage engine. Designed real-time monitoring dashboards with proactive low/high stock alert mechanisms for dynamic SMT operations.

Achieved complete end-to-end unit-level traceability. Optimized inventory levels across models and achieved significantly faster decision-making for production planning.

Procurement and Marketing functions were bogged down by repetitive queries and manual extraction of content from massive volumes of advertisement creatives and ERP documents.

Automate large-scale data extraction and deploy a conversational AI layer to provide near real-time, self-service access to intelligent enterprise insights.

Built an AI pipeline using high-accuracy OCR and RAG-based LLMs. Created a contextual classification system to detect branding, pricing, and PO data automatically.

Reduced manual queries by 60–70% and drastically lowered processing effort. Enabled advanced analytics by converting visual visual data into structured, actionable insights.

High-speed production environments relied on manual inspection, leading to inconsistent quality gates, frequent missed defects, and high rework costs from delayed detection.

To implement a computer vision-based quality gate integrated directly into the production workflow to ensure 24/7 consistency and alert accuracy.

Developed a real-time anomaly detection engine using image-based deep learning. Enabled continuous model improvement by feeding production anomalies back into the training loop.

Significant reduction in defect leakage and escape. Lowered rework and scrap costs while achieving a much higher level of product quality consistency.

Disconnected operational functions led to reporting lags and massive manual effort in calculating OEE, MTTR, and MTBF, preventing proactive factory management.

To architect a unified digital layer that automatically calculates maintenance metrics and provides a "Plan vs Actual" tracking dashboard.

Deployed an automated MTTR & MTBF calculation engine. Integrated maintenance scheduling with production logs to create real-time performance visibility.

Achieved a dramatic reduction in manual reporting effort and significantly improved planning accuracy through real-time "Plan vs Actual" insights.

The enterprise procurement function was overwhelmed by high-volume, repetitive queries, leading to delayed response times and inefficient resource allocation.

Introduce an AI-powered intelligence layer to handle vendor and PO queries, focusing on improving user experience and operational ROI.

Integrated RAG-based AI with ERP data to provide context-aware responses. Focused on self-service automation for PO and invoice tracking.

Delivered a 60–70% reduction in manual queries and established near real-time response capabilities, significantly enhancing operational efficiency.

High-volume SMT lines suffered from frequent stock-outs and overstocking due to a lack of clarity on model-wise consumption and material phase-outs.

Build a line-level visibility dashboard to monitor inventory consumption and trigger proactive alerts for production continuity.

Digitized shop-floor inventory and mapped consumption model-wise. Designed a proactive alert engine for low/high stock levels.

Optimized inventory levels across models and provided end-to-end visibility of inventory movement, ensuring consistent production flow.

Heavy industrial units were crippled by a total dependency on manual logsheet entries, resulting in high latency, data inaccuracy, and zero historical visibility.

To engineer a "Zero-Touch" data capture environment by integrating directly with PLC systems to automate machine-level logging.

Developed a direct PLC-to-Cloud synchronization bridge. Built automated stop-detection logic and heartbeat-based machine monitoring without operator input.

Achieved 100% elimination of manual logsheet entry, providing a reliable, audit-ready historical dataset and near-instant response to stoppages.

Managing large volumes of advertisement creatives manually was slow and error-prone, limiting the ability to scale competitive marketing analytics.

Build a scalable AI-driven pipeline to convert high-volume visual data into structured, machine-readable insights.

Architected an automated ingestion pipeline using OCR and Logo detection. Integrated contextual classification models for pricing and branding extraction.

Successfully automated large-scale extraction, providing consistent and accurate structured data for advanced marketing intelligence platforms.

Initial visual quality models faced accuracy drift over time as production environments changed, leading to potential defect escapes.

To establish a continuous learning loop where new anomaly data is used to improve model precision without disrupting the production line.

Implemented an image-based anomaly capture system that feeds "edge-cases" back into the model training pipeline for automated periodic updates.

Drastically lowered rework costs by maintaining high-precision inspection gates that adapt to line variations in real-time.