EPAM SYSTEMS INC Mi FID II RTS 27

EPAM SYSTEMS, INC. Mi. FID II RTS 27 and RTS 28 compliance May 2016

PROBLEM STATEMENT HEADCOUNT What is Mi. FID II directive for 1 FINANCIALS Mi. FID II will increase market transparency, efficiency and safety to the majority of non-equity products. One of the key Mi. FID II rules introduces stronger investor protection via improved disclosure to strengthen the best execution regime. Best execution regime 2 Challenging deadline 3 EPAM helped many companies to comply with best execution practice, including compliance with rules established by Securities Exchange Commission and FINRA. Specifically designed processes and analytics to generate reports for SEC Rules 605 and 606 Disclosures. Mi. FID II rules shall apply from 3 January 2018 2

The Markets in Financial Instruments Directive (MIFID II) KEY CONTROLS 3

What is in scope of Mi. FID II Best execution Mi. FID II extends regulation to listed derivatives, corporate and government bonds and warrants as well as structured finance products , emissions allowances , CFDs. Derivatives Corporate bonds Emissions allowances Structured finance products 4

Extraterritorial implications of Mi. FID II Implementation Delay Needs to Look Like 1)Many aspects of Mi. FID II will apply to non-EU firms, non EU-firms executing orders for their EU-clients, or executing in dual-listed financial instruments. 2)Asset managers that run global businesses broad plans to create compliance policies and implementation workflows. 3) Market participants that run domestic, non-EU investment may also have obligations coming from their EU-based customers. 4)The compliance departments of many EU-based firms are contemplating asking for changes in how their non-EU counterparties do business with them, in order for them to be compliant with Mi. FID II rules. A report into US equity trading carried out by TABB Group found 66% of buy-side firms will be impacted by Mi. FID II. 5

Non EU firms obligations under Mi. FID II 6

Mi. FID II BEST EXECUTION CATEGORIES COST SPEED PRICE BEST EXECUTION CATEGORIES NATURE OF ORDER EXECUTION & SETTLEMENT PROBABILITY OF EXECUTION / LIQUIDITY 7

NEW CHALLENGES TO ADDRESS : - Data challenges Sequencing events Time stamping accuracy Demonstrability Client facing documentation Broker/execution venue selection Access to ARMs Real time events reporting Many others 8

Mi. FID II Program management (c) GFT Mi. FID materials 9

EPAM’s SERVICES FOR EXECUTION VENUES and INVESTMENT FIRMS • Sourcing of data market data (FIX, binary) metrics of quality of execution from execution venues unstructured data (e-mail, IM i. e. Bloomberg chat parsing ) • Staging and aggregation data for analytics and comparison analyses for reporting for calculating volumes of instruments trading to control exemption and thresholds per instrument archiving, records keeping providing data architecture for firm’s orders and trades (supporting daisy chain) and market data 10

EPAM’s SERVICES FOR EXECUTION VENUES and INVESTMENT FIRMS • Analytics to calculate quality of executions on continuing basis calculation of effective and realized spread, probability of order execution, relative importance of factors such as speed, price, cost, likelihood of execution tool to generate reports on quality to regulatory agencies tool to analyze quality of execution tool to re-create market situations in the past , explore available liquidity at the time or order execution history • Digital design for dissemination and presentation of quality of execution information intraday and daily to clients via multiple channels clients’ facing interface (Web interface) 11

EPAM’s SERVICES FOR EXECUTION VENUES and INVESTMENT FIRMS • Real time and near real-time reporting for pre-trade activity (RFQs) trades Leveraging EPAM IP in market data protocols domain • API architecture to support UI and querying 12

APPROACH CONFIDENTIAL 13

HIGH LEVEL WORKFLOW Historical Market Information Historical Trade Information Intraday Trade Information Security Master: Rules of data aggregation across MIFID II affected asset classes Public Information: Performance of other venues trading the same instruments Mi. FID II processor Emails / IMs Collect order and trade information Data CSV / HTML Regulators / Data consumers OMS / EMS Collect market data information Quality of execution feedback Archiving / Recordkeeping Facility Aggregate information across multiple dimensions near real time Limits monitoring / Compliance Personnel Order / Trade / Market Data Repository 14

REVIEW OF EXECUTION QUALITY: WORKFLOW 1 • By clicking the trade dot user can see the all details about selected trade (order) 15

REVIEW OF EXECUTION QUALITY: WORKFLOW 1 • User is able to search trades by market place or by market participant or filter trades by specific parameters 16

REVIEW OF EXECUTION QUALITY: SUSPICIOUS TRADES 17

MARKET RECONSTRUCTION • User is able to review the order events and trades for any specific moment • User is able to switch to market reconstruction mode and move forward / backward with a chosen granularity in time using navigation controls • User is able to see order books for t-1, t and t+1 moments 18

REPORTING / ANALYSES. BONDS • Liquid and illiquid bonds Pre-trade and post-trade SSTI and LIS thresholds will be calculated in relation to percentiles in terms of trade count • Liquid bonds Pre-trade transparency requirements: the details of the transaction to be published (the price, volume and time of transactions) • Illiquid bonds A waiver / deferral may be granted because there is not a liquid market • Deferral period Two trading days after the transaction has taken place (‘T+2’) • Post-trade transparency requirements • Market operators and investment firms operating a trading should report: non-equity instruments which are traded on a trading venue • Investment firms, including SIs, which, either on own account or on behalf of clients, conclude transactions should report trades: outside trading venues (RMs, MTFs and OTFs) 19

BEST EXECUTION REPORTS. EXECUTION SPEED • Execution Speed Report analytics on average time between order submission and its execution Q 4 Summary: Avg Total Num Duration Average Execution Speed 0. 04 sec Platform Availability 99. 97% 1. 766 Consistency (Std Dev sec) sec Trading Interruptions 2 6. 23 sec Suspensions 2 5. 2 sec Average Order Acknowledgement Statistics for 2015: Chart View 0 -9 sec 10 -29 sec 30+ sec Q 4 99. 20% 0. 80% 0% Q 3 99. 50% 0% Q 2 99. 45% 0. 55% 0% Q 1 99. 30% 0. 70% 0% 20

BEST EXECUTION REPORTS. PRICE IMPROVEMENT AND TRADE EXECUTION QUALITY • Q 4 Summary: 2015 Statistics on Price improvement: Price-improved shares 89. 90% Q 4 Q 3 Q 2 Q 1 • Trade execution quality Platform Summary Report provides key statistics describing trade execution quality % Amount Industry Provider 89. 00% 94. 02% 1. 75$ 1. 99$ 86. 35% 93. 55% 1. 67$ 1. 83$ 86. 03% 93. 00% 1. 53$ 1. 77$ 85. 98% 92. 87% 1. 46$ 1. 64$ Q 4 2015 Trade Execution Quality Summary: At Or Better Than NBBO Price Improvement, % Execution Price (% of shares within NBBO) Execution Speed Effective/Quoted Spread Failed Transactions 97. 45% 93. 09% 95. 70% 0. 10 sec 42. 14% 0. 001% 2015 Platform Stability: Platform Availability Consistency (Std Dev sec) 99. 97% 1. 766 sec Outages Tading Interruptions Total Number of Suspensions Total Num 1 2 2 Avg Duration 1. 1 sec 6. 23 sec 5. 2 sec 21

BEST EXECUTION REPORTS. EFFECTIVE SPREAD • Q 4 Summary Platform Effective Spread 0. 0095 Average Volume @ BBO 195. 00 Average Spread @ BBO 0. 0095 2015 Detailed Statistics Average Effective Spread Industry Provider Average Volume @BBO Industry Provider Average Spread @ BBO Industry Provider Q 4 0. 011 0. 0095 20000 19500 0. 011 0. 0095 Q 3 0. 012 0. 0099 22000 21200 0. 012 0. 0099 Q 2 0. 0125 0. 0091 19000 17000 0. 0125 0. 0091 Q 1 0. 013 0. 0093 21500 20800 0. 013 0. 0093 22

BEST EXECUTION REPORTS. EXECUTION PRICE & LIKELIHOOD • 2015, Q 4 Industry Average Provider Shares executed @ BBO, % 96. 10% 96. 05% Total value of all offered rebates, discounts or other payments, % 5. 00% 4. 50% Total value of all costs, % 5. 50% 4. 99% 09: 30 11: 30 Average Executed Volume Small Medium Large 20000 2100000 3 mln 19999 2100000 2. 9 mln Time (T) Statistics: 09: 30 Industry Provider 13: 30 15: 30 Book Depth within 3 px increments 5 5 11: 30 Industry Provider 13: 30 21000 20000 2110005 3. 1 mln 2. 94 mln 4 4 Industry Provider 21030 20020 2121000 2115000 2. 99 mln 2. 65 mln 4 4 15: 30 Industry Provider 22500 21900 2300000 2328000 3. 5 mln 3. 45 mln 5 5 Description: * Percentage of shares executed at or within the National Best Bid or Offer (NBBO), which is a consolidated quote representing the highest bid and lowest offer for a security across all exchanges and/or market makers. It is updated continuously during market hours. * Total value of all offered rebates, discounts or other payments - is % of total traded value during the reporting period * Total value of all costs - % of total traded value during the reporting period volume * Time (T) Statistics - is obtained from price information reported in accordance with Article 4(1)(a) 23

Technology solutions for Mi. FID II challenges • MIFID 2 implementation deadlines: needs fast implementation options • Pre-built layers, including fast data load, transformation and visualization – outof-the-box available standard set of layers tailored for financial data • Cloud capable, sparing infrastructure procurement, setup of operation processes and teams • Enterprise security and operation: ready for governance challenges • Solutions for • • Confidentiality (security, access control, encryption) Availability (meeting SLAs) Integrity (permissions, data lineage, audit logging) Fit for changing needs during system lifecycle: small to large data for small to global enterprises • Scalable for the need of the organization and for the data volume and sources managed by the organization • Deployable with Dev. Ops support: scaling from a laptop as a server to thousands of nodes on a global private cloud data center • Long-term, cost-effective maintainability: modern and future-proof technologies • Long-term support for long-term operation - open-source tools backed up with enterprise support • Vendor independent - open-source solution for maintenance flexibility • Platform independent – support of multiple systems and cloud providers 24

Mi. FID II ACCELERATOR – SCALABLE BIG DATA DISCOVERY SYSTEM S 2 S 3 Data storage Data visualization and discovery Security & control S 1 Message queue Data processing • Cost-effective, scalable longterm storage with real-time reporting and data discovery features for financial data • Open-source solution • Dev. Ops: – Redundant, scalable components on each architecture levels – High maintainability as of modular setup – Standalone components ensure no data loss during maintenance – Automated (cloud) deployment procedure available Data providers / sources KEY TECHNICAL HIGHLIGHTS Zeppelin Active Directory Auth. Java API & auth. mid-layer UI and D 3. JS S 4 25

Mi. FID II ACCELERATOR – SCALABLE BIG DATA DISCOVERY SYSTEM Data storage S 1 S 2 Data visualization and discovery Security & control • Providing high velocity data flow • Sending messages with varied frequency based on business flow • Different data providers share different message types and formats (structured, semistructured, unstructured) • Message types can range from market to social data Message queue Data processing Data providers / sources KEY TECHNICAL HIGHLIGHTS Zeppelin Active Directory Auth. Java API & S 3 UI auth. mid-layer and D 3. JS S 4 26

Mi. FID II ACCELERATOR – SCALABLE BIG DATA DISCOVERY SYSTEM Data storage S 1 S 2 Data visualization and discovery Security & control • High throughput messaging system capable to store various messages • Elastic, scalable, expandable for increased capacity and performance with no downtime • Durable: messages are persisted and replicated • Persists messages while other components are undergoing maintenance Message queue Data processing Apache Kafka message queue cluster Data providers / sources KEY TECHNICAL HIGHLIGHTS Zeppelin Active Directory Auth. Java API & S 3 UI auth. mid-layer and D 3. JS S 4 27

Mi. FID II ACCELERATOR – SCALABLE BIG DATA DISCOVERY SYSTEM Data storage Data visualization and discovery Security & control • High performant data processor with in-memory analytic capabilities • Unifying messages to Hadoop and Elasticsearch in near real-time • Pushes processed messages to real-time dashboards using Elastic. Search • Ensures data discovery with rich visualization via Apache Zeppelin Message queue Data processing Apache Spark Data providers / sources KEY TECHNICAL HIGHLIGHTS S 1 Zeppelin Active Directory Auth. S 2 Java S 3 auth. mid-layer API & UI and D 3. JS S 4 28

Mi. FID II ACCELERATOR – SCALABLE BIG DATA DISCOVERY SYSTEM Data storage S 1 S 2 S 3 Data visualization and discovery Security & control • Cost-effective, scalable, fault-tolerant storage to host all current and historic data in a costeffective, redundant fashion • Integration with Spark for analyzing all datasets • Serving on-demand data pulls towards Elastic. Search indexes and 3 rd party systems Message queue Data processing Hadoop data lake Data providers / sources KEY TECHNICAL HIGHLIGHTS Zeppelin Active Directory Auth. API & UI Java auth. mid-layer and D 3. JS S 4 29

MIFID 2 ACCELERATOR – SCALABLE BIG DATA DISCOVERY SYSTEM • Ideal store for time series data • Clustered, scalable, highly available search system Data storage S 1 S 2 • Providing real-time search and advanced analytics for documents S 3 • Stores schema-free documents S 4 Data visualization and discovery Security & control • Manages current financial dataset for real-time analytics Message queue Data processing Elastic. Search Data providers / sources KEY TECHNICAL HIGHLIGHTS Zeppelin Active Directory Auth. API & UI Java auth. mid-layer and D 3. JS • Manages multiple indexes, including ones for ondemand data load 30

Mi. FID II ACCELERATOR – SCALABLE BIG DATA DISCOVERY SYSTEM S 1 Data storage Data visualization and discovery Security & control • Providing enterprise grade security • Integrating with corporate authentication systems, e. g. Microsoft Active Directory • Managing access permissions for data integrity • Maintains audit logs for data lineage • Provides secure interfaces for 3 rd party systems Message queue Data processing Data Governance - Middle layer Data providers / sources KEY TECHNICAL HIGHLIGHTS Zeppelin Active Directory Auth. UI S 2 S 3 Java API & auth. mid-layer and D 3. JS S 4 31

Mi. FID II ACCELERATOR – SCALABLE BIG DATA DISCOVERY SYSTEM Data storage S 1 S 2 Data visualization and discovery Security & control • Easy to use web interface for data analysts and scientists – using Spark in-memory computing engine in the background • Ready made visualization for ad-hoc queries • Exportable and shareable dashboards • Data science notebooks with Scala or (Spark)SQL Message queue Data processing Data discovery - Zeppelin Data providers / sources KEY TECHNICAL HIGHLIGHTS Zeppelin Active Directory Auth. API & UI Java S 3 auth. mid-layer and D 3. JS S 4 32

Mi. FID II ACCELERATOR – SCALABLE BIG DATA DISCOVERY SYSTEM Data storage S 1 S 2 S 3 Data visualization and discovery Security & control • Angular. JS and D 3. JS based visualization in a web browser – capable to plot large data volumes • Integrating Kibana based out-of-the-box available widgets for financial time series data boosting development performance and providing seamless Elastic. Search integration Message queue Data processing JS based user interface Data providers / sources KEY TECHNICAL HIGHLIGHTS Zeppelin Active Directory Auth. API & UI Java auth. mid-layer and D 3. JS S 4 33