Here you will find Case studies and White papers on various technologies.
Invoice Optical Character Recognition
The client is a major player in accounting software and supports its customers with invoice processing, accounting, and legal documentation. Most of the activities are automated but activities like invoice entry into the system are done manually for many vendors. Clients aspects to have automation models with error rates far lesser than human error.
AI BASED BLACK SPOT DETECTION FROM IMAGE
An AI tool has been developed to detect the presence of garbage, cluttered drain, unclean public toilets (known as black spots) in an image. The tool is intended to reduce about 80% of human intervention by automatically classifying images taken from various public places.
Technology Used: Tensorflow, Python, Android (JAVA)
Retail Checkout Automation
This product development consists of features that enable an automatic checkout (ACO) at the POS counter with minimal human intervention. Clients need our help in developing recognition models for 1000+ products and deploy an optimized algorithm on the POS edge system. Fallback parameters were defined for the best-optimized model and minimum compute resource requirements.
Automotive, Resale Price Prediction
The company situated in the US, primarily into Bike and ATV resale retail + wholesale business recently diversified into car retail resale business via acquisition and also launch of a new online platform for buying and selling cars. Requisitions for car sales would come online and this requisition was married with CARFAX and Manheim Data for Inspection. With close to 1500+ requisitions per day, the inspection staff was under pressure to provide buy price quotes for all requisitions. Over prediction means a loss to the company and under prediction means loss of opportunity. As a result, profitability suffered, Wall Street gave a low rating, and share price plunged.
CPG, Trade Promotion
Major CPG goods manufacturer in the UK wants to know the effectiveness of a certain kind of trade promotion ( on the shelf, a display coupled with discounts, volume promotions, etc.) and then use this information to maximize its business metric ( Market Share, GMV, Revenue, etc.) within multiple constraints of budget, price, target revenue, etc. The client was earlier using rudimentary methods with information sourced from external data vendors to understand the effectiveness of promotion and then use business heuristics to optimize.
One of the biggest electronics retailers in Dubai operates 30+ stores and offers 15000+ different SKUs at competitive prices. The client prepares for the annual shopping festival/season which is the major period of demand and also depends upon ~20 categories (~500 SKUs) for most of its revenues. The shopping period is planned with many offers, discounts, and activities to attract potential customers. The client is currently doing forecasting at a very aggregate level which is category/quarter which is finalized via consensus on-base forecast. Execution on a granular level is owned by store/category owners and is mostly subjective.
AI based system to measure the calorie burned during daily activities
The clients were prestigious, and they had experience in sports gear manufacturing and smartwatches for over a decade. They had been using the advanced rule-based system to read sensor values from the smartwatch and then predicting the activity of the user. This rendered them into a situation where they had to compromise with their performance because real-life scenarios were often unpredictable. The utter reliance on the smarter system was mandatory and hence the requirement of AI integration was absolute. They wanted to emulate more practical scenarios using nonlinear and non-deterministic systems.
ASR- AUTOMATIC SPEECH RECOGNITION
Automatic Speech Recognition (ASR) has one of the main goals to allow oral communication between humans and electronic devices (computers, mobile, etc.) i.e. consist of the process of converting the human speech into textual representation automatically. The main problem of the ASR is the complexity of the human language.
Multi Camera People Tracking
Popular university in the UK with a campus spread across 300 acres, offering 25+ Programs in Graduation and Post Graduation. Has 150+ facilities to monitor and maintain wants a deeper peek into student behavior. At any given point in time along with the classes, there will be at least two dozen small and big scheduled events running on the campus along with seasonal admissions into the programs.
Image Categorization: E-Commerce India
Indian fashion E-Commerce giant wants to use Machine Learning for competitive advantage by understanding competitor catalog and take decisions on pricing and catalog upgrade. This exercise also has to do with events, festivals, and promotions. The overall aim was to be ahead of the competition in terms of merchandise and pricing. The outcome will be seen in terms of customer experience and gross merchandise value.
Retail, Shelf Space Optimization
Company in India which has exclusive rights to market and sell premium European brand of stockings and sportswear. The company places 150+ turner shelves across 50+ stores across India. A store is either a store in a mall or a supermarket/hypermarket while a turner is a turning shelf like a vertical platform where a fixed number of units can be placed in 24 to 40 shelves depending on the type of Turner. The company makes a fair amount of profit but operational inefficiencies are piling up as material once bought cannot be returned back to the parent brand, inventory balancing cannot be done due to geographic spread because of which volumes will not justify the cost but local stores can return to the client.
Airline, MRO Inventory Optimization
Singapore based MRO company wants to rationalize the inventory of Spare parts to minimize overstocking carrying cost, CAPEX cost, and penalties due to understocking. MROs are service partners to airlines who do not want to have their specific service facilities or if owned service facility is not available at a certain airport or for a specific kind of service need a partner. Inventory is categorized as critical or non-critical. Part value can range from a few dollars to thousands of dollars as well. SLA of service varies from client to client and thus penalties in care of no service also vary.
Aviation, Parts Pricing
The client is among the largest distributors of aircraft parts in the US. Wants an intelligent pricing model for its quotations process. There is no background data available about competitors or probable customers. A lot of data suggest that competitors had been doing window shopping to understand their pricing strategy. Overall 40000+ SKU with a lot of overlap as many parts can fit the same requirement.
Implement FIFO for 3PL
Leading Indian 3PL company manages organic perishable goods warehousing and mid-mile distribution. The client handles One Lakh SqFt of warehousing space and average daily distribution of 50 tonnes in three shifts of 24 X 7 operations. 3PL contract demands set limits on wastage and to handle this scenario client wants to implement processes in quality check and FIFO for material handling.
Process Mining and
Dashboarding for E-Commerce
E-Commerce leader in India handles three million-plus shipments per day across its complex Supply Chain network of First Mile, Forwarding Centers, and List Mile hubs across India. Has implemented state-of-the-art automation and data capturing systems to regulate and capture the flow of the shipments as they move along each process. Our client wants to use this huge amount of data produced daily in improving order SLA and RCA cause of the delivery breach.