Unicsoft

Business domains: Trading

Expertise: Blockchain, smart property & smart contracts

Technology and tools: Angular.JS, Python, Django, Blockchain API library 1.3.3 & 1.4.0, MLlib, pandas. Docker

The goal of the project was to create a supply chain transparency platform capable of tracking each piece of gold electronically and in a secure manner. At the time there already were similar successful cases in other industries, for example, jewellery and medical implant tracking, which were Blockchain based solutions, which gave us insight regarding the technical solution in this case.

The basic technical complexity was that existing platforms track the origin of one product, for example, an implant. With gold, it is more difficult – the mine produces large ingots of unrefined gold, further refineries clean it and one unrefined gold ingot produces several bars of a smaller size of high-grade gold, then other institutions can melt these slivers into smaller ones, so the original “product” is crushed into components and it is necessary to track the way of each component back to that big ingot from which it went. Thus, the ready solutions were inapplicable in this case.

The problem was solved by applying the QR code to each ingot and to the derivatives of ingots and creating a custom system for registering all operations of the type in blockage (ingot A was turned into ingots B, C, D by X LTD, ingot A itself no longer exists) and all displacements (ingot B is moved from companies X to company Y with the help of transport company Z).

The system is realized with the help of Ethereum, and the logic of operations with ingots is programmed in the form of smart contracts.

The MVP was approved by the National Mining Association and (most likely) will receive investments from this organization.

Business domains: Finance, e-Commerce

Expertise: Cryptocurrency, Blockchain, Finance-related tools

Technology and tools: Go, Angular 2, Blockchain

This request constituted of creating a cryptocurrency and embrace it in environment containing related service. The Client was a Chinese company that was striving to launch its own cryptocurrency for economic purposes and to increase its investment-oriented attractiveness, and, last but not least, – to successfully complete pre-ICO and ICO stages.

The Client’s roadmap contained the development of the coin itself, cryptowallet, and also featured bespoke framework for Consensus and Opinion over p2p networks.

Development team was distributed and consisted of ten software engineers located all around the globe. Go was picked as the main language, so all of the tools and environmental components were written using its native tools. It’s worth mentioning that one of the peculiar parts of project was an implementation of Merkle tree. Main concept of the tree might be described as a tree where, every leaf node is labelled with the hash of a data block and every non-leaf node is labelled with the cryptographic hash of the labels of its child nodes. This approach enhanced security and reliability level of the entire transactions system.

After successfully getting through ICO stage the cryptocurrency was launched into market in September, 2017. Project is currently under ongoing development and is being geared up with new features. The Client remarked high level of Unicsoft technical capabilities and versatility of management approach.

Business domains: Media, Social

Expertise: Behavioral analytics, Predictive analytics

Technology and tools: Python, SQL & noSQl, R, mlLib, Spark

One of the most renowned EU&USA publishers reached out to us with an intention to build a new value-added service for their subscribers focused on prediction of a bill-voting outcome in US congress. Such service was offering an innovative, impartial and accurate approach to political analytics, as opposite to classic “expert” approach used by other players on the publishing market.

The journey began with deep analysis of the problem and available data to solve it. The main challenge was that in order to get high accuracy the machine learning model had to gather attributes of a bill and voting habits of senators from multiple non-congruent sources. Added complexity was the fact that both “bill” and “voter” are complex entities and specific voter decision regarding specific bill depends on multiple attributes and their combinations, therefore requiring advanced approaches for prediction.

Long story short – the model had an accuracy of 84% proving the business case. Such successful proof of concept initiated series of new initiatives and value-added services based on the data existing in the organization and utilizing the power of data science and machine learning to gain unique insights from it. This solution became a capstone for a versatile solution allowing users splice&merge datasets and predicting not only bill-voting outcome but a precursors leading to the decision of even factors, affecting behavioral pattern of a particular congress member.

Business domains: Finance, e-Commerce

Expertise: Machine Learning, Data Security

Technology and tools: R, Python, mlLIB

We live in the “era of data”. Everyday companies face a compelling number of interactions with data source, and the more data is being produced, the more risks of this data to be used maliciously arise. Our Client, European start-up had an idea of developing a solution for the forecasting of spear-phishing of personal data for a wide scope of organizations located in various countries. This type of solution is excessively sought-after nowadays – especially combined with ML approach for striking prediction accuracy. Knowing about Unicsoft expertise in that domain and having previous successful track record with us, Client decided to choose us as an authorized solution developer.

Firstly we defined main datasources: the pinnacle was VCDB database (catalog of data security incidents using VERIS framework). Second source used in the analysis is FT500 data set in 2016. Then, after merging these two data sets we obtained set for predicting data breach probability for a particular company. Key ML models were GLM and Random Forest models with RF prevailing because of its higher precision. Moreover, after implementation of Monte-Carlo simulation methods we added prediction of incident density within particular timeframe. As an additional perk we predicted expected loss in USD in case of attack for various industries.

Solution was accepted positively right after its demonstration by various investors and got a lot of encouragement for the further development into more complex product with extended functionality. Due to model precision that amounted to 83% and the fact of Monte – Carlo simulation techniques implementation one may be confident that abovementioned solution has a huge potential and versatility.

Business domains: Finance, e-Commerce

Expertise: Blockchain, Smartcontracts, Tokens

Technology and tools: Back-end: Silex, mySQl, Redis (API websockets); Front-end: Angular 2 Blockchain: Parity + Node.js

European fast developing company specialized in online retail of goods for children reached out to us with a request to create a factoring platform for their supply chain. Since classic factoring business is prone to high operations costs and slow factoring requests processing, the “classic” was not aligned with company’s agile and innovative culture. The solution was to apply Blockchain, which offered an automated low-overhead way of securing the transactions and enabling the use of smart-contracts technology. Main intention lay in gaining a competitive advantage due to development of a solution with use of a top-notch technology leading to the highest security and reliability.

As a result of business analysis project team came up with the high level technical solution which included combination of Ethereum and Parity allowing getting benefits of both modern and lightweight Proof-of-Authority mechanism and SmartContracts.

Back-end of the tool had two main nodes – first one was kind of admin-panel, and the second one was public. Admin-node was intended to provide a synchronization of all invoices that were stored in the tokens using internal API. Internal API, in its turn, reached to the Smart-Contract module enabling use of abovementioned technology. Public node let users get any requested information regarding the tokens. Initial infrastructure was deployed using AWS because of reliability and resilience of that service. The development team consisted of one front-end developer, one back-end developer, and one designer. Overall timeframe amounted to eight months.

Business domains: Finance, Microfinance

Expertise: BBlockchain, Predictive analytics

Technology and tools: Angular.js, Python, Django, Blockchain API library 1.3.3 & 1.4.0, MLlib, pandas. Docker

Client decided to extend their audience due to to novelty and security of the product. One of the peer-to-peer lending platforms, operating in the UK & Ireland and regaled as an one of the leaders in industry asked Unicsoft to provide development services. The key peculiarity of aforementioned system was the use of Blockhain technology. Client wanted to implement smart-contract system that would allow Users to exchange secured contracts alongside with opportunity to request & receive loans in Ethereum as well as in more “accustomed” currencies.

One should mention that all of the Blockchain-related frameworks were successfully integrated to the system, particularly, to its core. As far as an adherence to microservice paradigm was chosen, we were quite indubious regarding framework for UI, and perceiving both compatibility and resilience we chose Django.

Then a full-scale platform was developed within 12 months work.

The platform was successfully launched and reached its breakeven point within 6 months, providing the Client with the growing Users portfolio and maintenance of financial risks thus positively affecting the revenue. Portfolio, in its turn, accreted to 115% within three months after launch compared to a same period of the previous year.

Business domains: FMCG, brand management, customer response prediction

Expertise: Bigdata Processing and Analysis, Advanced Analytics

Technology and tools: R, Python, Tableau

One of the FMCG leaders, Large tobacco company, approached us with a need to expand their brand-line alongside with increasing the number of distribution outlets. The Initial and foremost goal was specified as prediction of customer response to various marketing collaterals and design elements, i.e. different advertisement types, external branding elements such as outlets, big boards, and even colour combinations used in visual materials. Large tobacco company possessed a significant amount of historical data comprised of sales data and description of branding action of a particular moment. The final model was supposed to be complex and intended to detect any response within the predicted deviation of customer’s behaviour or purchasing activity which might have been caused by promotional and/or advertising activities.

Having considered the Client’s current and future needs, the key points of the solution were defined as: fine-tuning of current Client ETL solution; multicomponent predictive model containing several sub-models; price elasticity and sensitivity model; embedded Bi-solution showing customer’s response to any applied activities in both short-term and long-term time frames.

The initial effect of the model became noticeable and appreciable within first four months. In addition to that, due to the complex analytical model for new activities prediction and doing business considering its “advice”, Large tobacco company managed to have a solid increase through customer response rate and, consequently, a revenue increase.

Large tobacco company pointed out a remarkable level of Delivery as well as high level of overall performance provided by deployed solution.

Business domains: FMCG, commerce/retail

Expertise: NLP, Text Mining, Data Science

Technology and tools: Python, NLTK, Apache Solr

The largest Central & Eastern Europe retailer asked Unicsoft faced a need of a tool that would be a capstone of category management: having over 50000+ of PLUs and getting a dozens of new names every couple of days, Client had only a manual tool for category definition. So far, the basic intention of a tool was a creation of “tag clouds” supposed to tell one sub-categories from the other ones based primarily on a descriptive information, and, consequently, for proper management of PLU turnover, placement and promotion.

First step was defined as data preparation and enrichment: given more than 50k PLU names alongside with variety of features (main category name, manual tags, supplier notes, etc.), a main thing was to set up an proper DWH solution for storage and handling with further data pre-processing activities. Starting with SQL-type DB combined with Python libraries for pre-processing, a huge bunches of data was cleaned up and enriched. Next step was a tokenization of data: main goal was to create a tags cloud for working out subcategories : I.e, given the “sweets” category through applying these methods based on keywords, a list of sub-categories came up (diabetic sweets, soya sweets, etc.). Subsequently, that approach was applied to entire portfolio of PLUs providing opportunity not only to categorize existing goods but to process incoming ones through fetching their description and its analysis.

Break-even point for this particular solution was reached after 9 months of use through saving a significant amount of money by the cost-cutting on the staff that previously did such activities manually.

Business domains: E-commerce

Expertise: NLP, Text Mining, Data Science

Technology and tools: Python, NLTK, NLP

One of well-known Central Europe e-commerce merchants asked Unicsoft for cooperation to create a tool intended to analyse a customer’s feedback on goods purchased through web marketplace, The business goal was to increase customer loyalty, drive business changes, and deliver real return on investment. Customer experiences fall into three basic categories, positive, negative or neutral. Through sentiment analysis, the goal was to detect the tone and temperament of each and every word found in a customer’s social postings and categorize those sentiments as either positive, negative or neutral.

After data cleaning and munging , an initial step of words tokenization was applied; having it done, an NLTK tools was used to define synonyms, semantics, and overall mood of feedbacks; aligned with scores given by these particular feedback authors, a section of manual business logics was brought in: language specifics, abbreviation, collocations and vernacular expression played a significant role in overall semantic analysis. Furthermore, analysis of new products or product lines could significantly affect overall strategy. Alongside with NLP and semantic analysis, there were used a Data Science techniques : given a data from social network that customer logged in through, a set of demographic features was involved to model, summing up into complex analytical solution.

Unicsoft set a dedicated team comprised of highly-skilled analysts and mathematicians and software developers; this particular solution helped the Client to define his marketing and sales strategy which resulted in 10% revenue increase within one year after the deployment. The Client noticed the overall high-level of Delivery and solution architecture.

Business domains: Automotive, Advertising

Expertise:Bigdata Processing, Predictive & Advanced Analytics, Machine Learning, IoT

Technology and tools: Apache Hadoop, Hive, Cassandra, Python, Scala, mlLib

One of the Leading Automotive Dealers (further – AD) that has been performing its activities since the 1990s in CIS and Eastern Europe is looking to increase the company’s revenue by exploring new monetization opportunities. The idea is to use the data collected from smart devices installed into a car for Target Advertising and to increase additional revenue stream through presenting precisely targeted, high-quality ads to car owners and sharing added value with selected DSP and SSP platforms.

AD’s internal analytics team required additional expertise which was not available at the moment, specifically: expertise in complex BigData processing and building SaaS solutions for IoT and Advertisement domains. Unicsoft was able to offer the required expertise which laid grounds for further long term collaboration.

Unicsoft successfully implemented and deployed the solution which allowed AD to build additional revenue stream in cooperation with several DSP and SSP platforms; R&D investment carried out by AD reached the breakeven point within 12 months.

The prediction of car owner’s response lead to lower than expected number of opt-outs, positive response of car owners to new service and high advertising conversion rates.

Given these projects are completed, AD indicated a high level of delivery, R&D value and top-notch class of specialists involved. AD encouraged Unicsoft on further partnership that led to another Data Science initiative in Automotive Insurance domain.

Business domains: Media, Entertainment, Recommendation engine

Expertise: Bigdata Processing and Analysis, Machine Learning, Predictive Analytics, Recommendation Engine

Technology and tools: Python, Scala, Apache Spark

One of the leading Software Development Agencies (SDA) from the UK contacted us for cooperation on behalf of one of its Clients, who performed activities in the Content Media Market. The end-client successfully launched a platform similar to 9gag and Reddit alongside a mobile application. The Client’s business was focused on enhancing platform monetization by increasing the end-user’s loyalty and willingness to click & go to the partners’ offers.

The high-level goal was to build a recommendation engine which would use Big-Data analysis and machine learning to analyze individual end-users’ preferences and to help discover new content accordingly.

The project comprised two substantial parts – setting up the dedicated team of experts for continuous collaboration and transforming business needs into added-value solutions, in addition to direct development of the aforementioned. Unicsoft allocated two in-house experts, as well as quickly engaged one pre-interviewed expert from the HR pool. The proposed solution consisted of researching and developing two different Recommending Engines, each of which used different approaches (content-based and user-based).

As a result, SDA met all the deadlines for this Client, having brought the highest level of delivery and technical expertise. Client, respectively, managed to increase its audience monetization crucially (for the first six months after the deployment, click-and-go for partners offers grew by 20% compared to the previous period), whilst the overall audience increased by approximately 8%.

Business domains: Gaming, E-commerce

Expertise:DWH, Big Data Processing, Machine Learning, AI, Fraud detection & prevention

Technology and tools: Apache Spark, Hadoop, Hive, Python, mllib, Random Forests

A well-known online game producer that launched the line of browser games faced a need to maximize monetization of one of his games through a thorough analysis of gamers’ data (purchases, activities, etc.) as well as need to make internal game security stronger due to up-to-the-minute detection of fraud and cheating activities.

As a result of Business Analysis stage, Unicsoft identified following deliverables as required to fulfil set business goals:

1. Gaming analytics: given a historical data for entire game users and after its preprocessing (cleansing, enrichment, etc.) a cluster of active users has been defined. Then, key features of gaming behavior and in-game purchase occurrences were carefully analysis with the purpose to tune up gaming AI: main thing was to keep balance between game difficulty for keeping gamer involved and adding several “hard spots” which might be passed easily if particular in-game purchase made.

2. Anti-fraud module: analysis of suspicious activities occurred during gameplay or internal currency purchase: detecting fake cards and adding them into database; detection and consequently prediction of cheating activities in game process (multiple accounts, fake teams, etc.)

3. DWH/ETL auditing and improvement: regardless of having plenty of various datasets, all of them were stored in old and rather unstable SQL-type solution, so need of rapid access arose eventually; after thorough analysis of requirements, Clients’ data was successfully transferred to BigData solution that consisted of Hadoop + Hive alongside with analytics modules plugged onto.

Business domains: Commerce, Pharmacy

Expertise: Data Science, BI, Predictive analytics

Technology and tools: Python, noSQL, xgBoost, Tableau

One of Unicsoft clients, a large pharmaceutical network came up with request for a collaboration regarding development and implementation of OOS (Out of Stock system). For that current moment, Client had over 2000 drugstores all around the country and a reporting module comprised of reports concerning revenues and daily turnover.

The main goal was to set up a proper ETL process and perform a predictive analytics module intended to predict absence of drugs & remedies alongside with BI-tool deployment for ad-hoc reports and predictive analytics. Subsequently, the request evolved into not only a analytics & reporting module, but a complex solution providing access for suppliers to let them see the picture considering sales & trends of their products in any point of sale.

Unicsoft started from setting a dedicated team comprised of data architects and engineers to establish and consequently maintain data migration process. Considering necessity of being able to do close to-real-time analytics as well as requirements of DB consistency and availability, a noSQL architecture was considered as the most suitable. Thence, a several data scientists were involved to manage data enrichment and munging process. Having this step completed, a predictive analytics module was implemented to calculate estimated risk of product absence and predicting sales trend. As a copestone, a BI module was added providing a visualization of predictive module outputs.

Client underlined well-organized and orchestrated delivery process from Unicsoft side, and, consequently, this case of collaboration led us to establishing a successful cooperation along with new projects for pharmaceutical industry.

Business domains: Finance, Insurance, Automotive

Expertise:Bigdata Processing and Analysis, Predictive Analytics, Machine Learning

Technology and tools: Python, XGBoost, Neural Networks, Decision Trees

Being aware of successful cooperation between Unicsoft and one of Lead Automotive Dealer, an International Insurance Broker (further referred as IB) contacted Unicsoft with an idea to develop a solution for a data-driven assessment. Being a partner of several automotive dealers, IB has an access to data received from in-car gauges and in-car smart devices. Therefore, there are two sources of available data: historical and descriptive data about users and the above mentioned data coming from devices, which build the foundation for increasing accuracy of the final model through determining a driving style and User’s proneness to an accident.

Key points of solution for IB were defined as: choice of proper solution for handling streams coming from two data sources (internal from IB and external from ADs); data cleaning, noise and outliers reduction; development a Scoring model for risk level and proneness assessment;machine learning and model improvement within a specified period of time after its launch.

Unicsoft set a team of Advanced Analytics experts along with DevOps and Data Warehousing specialists, which completed initial data cleaning and predictive model within two months and a deployment-ready version of solution within seven months; further, a support team was set up and introduced to IB.

The solution was successfully launched and had a commercial success with reaching the breakeven point within the first 18 months. The solution generated additional sales by offering a personal value proposition due to implementation of multi-feature risk assessment model and proper approach of rate evaluation for a particular User.