CRM Data Migration
Objective
To help client improve data migration speed. Client was experiencing rapid growth of subscriber numbers in their CRM platform. The problem was that the user would sign up only if their data could be migrated. The client was losing customers due to their inability to catch up with data migration speed.
Solution
Lifewood engineering team redesigned the workflow process with modular application that can be configured to handle different CRM sources with different formats.
Lifewood Project Management team also formulated a communication format to facilitate communication with clients, reducing errors and clarification needs.
Result
Improved data migration efficiency by 4 – 23 times.
Volume: ~23 migration per month
Turnaround Time: 2 days for each migration

Anomaly Detection of Legal Document
Objective
- To create a machine learning pipeline for anomaly detection applying to legal document in PDF format.
- Processes include digitization, text extraction, text interpretation, transcription, and structure them in spreadsheet.
- Multi-lingual (~53 different languages)
Solution
- Apply OCR technology with high accuracy
- Workflow with automated text extraction and formatting/segmentation with manual text verification.
- Multi-stage quality checks to achieve above 98% accuracy.
- Collaborate and mobilize teams from 15 different countries to handle 53 languages.
Result
- Successfully prepared 20K document for training data within 6 months.
- Met client target for 98% accuracy.
- Volume: ~130K pages

Information Extraction using Natural Language Processing
Objective
Automate information extraction from old newspapers on target components including. Date, Name, Medal, Rank, Group.
Solution
Full implementation of deep learning solutions using the latest transformer-based NLP library – Spacy. To handle the uniqueness and historical text format, the model was trained from scratch using more than 10,000 dataset.
Production workflow involved OCR, the relevant data are automatically extracted with Named Entity Recognition (NER)
Result
We integrate the AI information extraction automation to our LiFT platform and achieve 10-fold productivity improvement compared to common used methodology.

Multi-language Voice Collection & Transcription
Objective
To collect a total of 5K+ hours of voice recording from native speakers from 10 different countries: India, Malaysia, Indonesia, Philippines, Vietnam, Myanmar, Laos, Germany, France, and Russia

Solution
Recruit resources from 10 different countries in parallel to manage the uninterrupted incoming recording. We also deployed our own web and mobile app to meet client requirement and expectation
Result
Together with our industrialized workflow and strong network of our crowd team across the regions, we deliver the project ahead of schedule.

Real-Time Online Waybill Processing Services
Objective
To provide real-time online tracking waybill process services. The client is one of the largest express delivery company in the world.
All couriers are express orders sent out from China, written in English and in handwriting.
We provide 9.5 minutes/20 orders for time-effective documents and 15 minutes/20 orders for parcels.
The accuracy rate of important fields needs to be 100% correct.
Solution
To achieve data security and timeliness, We applied division of labour principal, data was processed by different operators and then merged and send back to customer.
For critical fields, third -party quality inspectors were employed to ensure data accuracy.
Result
For 15 consecutive years, we handle on average 30,000 Waybills per week, enabling the client to query waybill status online in real time.
Use the duplicated data entry process to ensure the accuracy of critical fields to 99.99%

Electronic Information Security Solution for Bank Paper Documents
Objective
To serve a bank with real-time processing of credit card application in paper forms. Turnaround Time is 1-4 hours depending on urgency of the applications.
High demand in Data Confidentiality. High demand in Data Entry Accuracy.
Solution
To ensure data security, we work within the client’s server environment. Our operators use Citrix to connect to the client’s server within authority limit.
Before data processing, the image is automatically desensitized (covering sensitive information) and cut into different portions. Individual data portion is assigned to different personnel to process and then return data back-to-back for data consolidation.
Result
For 10 consecutive years, we have processed about 10,000 electronic receipts every week.
Deploy duplicated data entry methodology and run the difference process to ensure that the data accuracy rate is up to 99.99%

Paper Invoice Digitalization solution
Objective
To process large quantities of invoices and related paper documents, helping customers to process financial settlements faster.
14 types of documents, including sub-invoices, equipment lists, time sheets, delivery lists, purchase orders, receipts etc.
The documents of 2,000 contractors and subcontractors, having different formats, including English, French, German and other minor languages.
The required accuracy of different fields and character levels ranges from 98% to 100*.
Solution
After analyzing the sample data, the images are sorted and classified before processing. According to the degree of difficulty after classification, tasks are dispatched to operators of different skill levels for processing.
Two-step verification of important fields.
Result
A 300-person team completed the digitalization project of 3,000,000 documents in 4 months.

Establishment of Business Background Check Database
Objective
To convert all information on the paper yellow pages into electronic format and confirm the data through call center. Establish database information came from various enterprises and institutions in China. The data source came from various yellow pages.
Challenge 1: Traditional data entry has low efficiency and low accuracy.
Challenge 2: The output data contains incorrect information, outdated information, and data accuracy cannot be guaranteed.
Solution
Process industrialization to high efficiency and quality, consistent result.
Dual OCR verification, automatic dialling.
Adopt end-to-end solution with scanning-digitization-phone verification.
Developed a customized call center system and professional customer service team for this project
Result
Provided services to customer for 4 consecutive years. Every year helped client to add 1,000,000+ records to their database.
Customer use our innovative process “Dual OCR & Manual Verification” to gain higher productivity and better quality for win-win achievement.

Debt Management Database
Objective
To service as back-office center for customer.
Maintain the Debt Management Database in a highly secure environment.
View and understand various types of documents, updated private personal financial information in the client system.
365 days covering 2 country sites
Solution
Background check for all operations.
Information security training twice a year.
Set up a five-layer physical security environment.
Operators work in a high-security room equipped with 24×7 CCTV.
Log in to the client system remotely through Citrix Remote Access.
Result
Work scoped expanded to include document classification enabling revenue increased by 50%
Zero security incident.
ISO 27001 certified.
Debt Management Database
