Content thumbnail MatLogica Pitch Deck
AI Content Chat (Beta) logo

MatLogica Pitch Deck

Future Tech | MatLogica is a B2B Deeptech/Fintech, that developed a breakthrough software solution that unlocks a full potential of modern CPUs

Unlocking full potential of modern hardware A UNIQUE COMPILER THAT SPEEDS UP REPETITIVE CALCULATIONS AND AUTOMATICALLY GENERATES DERIVATIVES 6-100X FASTER

LOGICA QUANT’S NIGHTMARE MATLOGICA - SOLUTION Intraday risks too slow? Risks not granular enough? Never-ending refactoring? Code becoming unreadable? Unresolved legacy bugs? Inaccurate P&L explain? Overnight batches too long? Exis % ng Quant Library MatLogica Exis % ng Hardware A layer of abstraction delivering: - Automated computation of sensitivities (AAD*) - Data-oriented performance for object-oriented code ✓ Speeds up: native CPU vectorisation + safe multithreading ✓ Precise analytical sensitivities ✓ Simplified development and faster time- to-market for new models ✓ Consistent risk methodology ✓ Grid/Cloud savings + Simple, semi-automated integration approach: initial results in weeks! * AAD = Automatic Adjoint Differentiation

MatLogica Copyright 2021 MATLOGICA HOW IT WORKS 3 OPTIONAL CREATION OF ADJOINT KERNEL Transform valuation graph into kernel X Y Z A Optimised valuation graph Xeon avx512 kernel Transform valuation graph into kernel Xeon avx512 adjoint kernel FunctionB() OO OO OO Start() FunctionA() FunctionC() FunctionD() Extract valuation graph using Operator Overloading The kernels can be sent to the cloud, enabling scalable and secure cloud approach!

MatLogica Copyright 2022 MATLOGICA A PARALLEL COMPUTING SUPERCHARGER 4 50% Lower Grid/Cloud costs Speed-ups of 6-100X For pricing & Risks 3-4X Faster IT turnaround A library that supercharges quant analytics and enables fast Automatic Adjoint Differentiation (=sensitivity/back-propagation), with only small changes to the code XVA Modelling Machine Learning Enabling AAD for CUDA Live Risk Sensitivities Curve Construction Exotic Pricing Scripting Languages ‘What if?’ Analysis Monte-Carlo simulation EPE (expected exposures) = Lower CO2

MatLogica Copyright 2022 MATLOGICA OUR RESULTS SO FAR 5 2022 : ‘Category Leader’ in XVA Components Quadrant 2023 : ‘Category Leader’ in: • Innovation; • AAD; • Data Parallel Programming; • Innovation in Computational Frameworks; ... ranked #10 out of 50 in the QuantTech rating ! Joint performance benchmarks with Intel - 1770x speed up for XVA pricing Client results: production-ready in 9 months Pricing and Greeks Pricing only 1 AVX 512 core 23x 35x 56 AVX 512 cores 831x 1770x Risks 15-20x faster; Overnight batch from 8 to 2 hrs; 70% quant code reduction ... and more! Among 15 start-ups in London Accenture Fintech Innovation Lab, cohort of 2023 4 paying customers: 2 banks & 2 academia; PoCs ongoing with several Tier 2 banks

LOGICA IT’S EXCITING BECAUSE THE POTENTIAL IS HUGE! Business Model Annual Licence Fee, charged per use case Licence cost - Tiered Approach (beach-head @ $100k, later up to $500K in ARR) 1+ Licence per Client Projections - by 2028 Best Case: Cross-Industry Leader (Finance, Energy & ML) - ARR £1B +; ROI 100x+ Realistic: Sizeable Presence in Finance & Energy Markets - ARR £200M ; ROI 20x+ Worst Case: Remains niche product - ARR £10M ; ROI 2-3x+ Market Size By 2028 $180B+ $120B+ $20B+ SOM TAM SAM Financial institutions with large trade portfolios, est. > 1000 globally Machine Learning + HPC ~15-20% of SAM

MatLogica Copyright 2022 MATLOGICA TOP CHOICE FOR SEAMLESS PERFORMANCE AND AAD RISKS 7 The only technology that speeds up simulations AND computes risks using AAD Doesn’t require learning new languages or extensive re-work Enables secure cloud: only binary kernel to be exposed, no models or portfolio information Allows to focus development resources on activities bringing value, not optimisation: low-level optimisations as vectorisation, multithreading and memory use are automatic CO2 savings via reduction of grid/cloud usage

MatLogica Copyright 2022 MATLOGICA AND OUR VISION IS MORE: LIVE RISK 8 ‘Always on’ Kernel Market data Live prices & Risks Traders

MatLogica Copyright 2022 MATLOGICA TOP CHOICE FOR SEAMLESS PERFORMANCE AND AAD RISKS 9 “Quants will find that Matlogica seamlessly fits with their libraries and effortlessly accelerates pricing and risk by impressive amounts." Antoine Savine, Author of “Modern Computational Finance”

MatLogica Copyright 2022 MATLOGICA OUR LEADERSHIP TEAM 10 Dmitri Goloubentsev Tim Watmough Natalija Karpichina Sebastian Steinfeld, MPhys and MSc, Quant Analyst Andrei Goloubentsev, Quantitative Analyst Antoine Savine, PhD in Mathematics, Author of Modern Computational Finance (Wiley, 2018) CTO SALES LEAD COO ADVISORS PARTNERS 17+ years working as a quant/IT Geek & Mathematician Owned Executive Search firms Long-term relationships with potential customers 10+ years working in Finance & Energy IT MSc of Software Engineering + MBA PUBLICATIONS

CONTACT - [email protected] WEBSITE - WWW.MATLOGICA.COM Looking for: SUMMARY Appendices: • Testimonials • Case Study • Competition • Geeky notes • Technical Implementation MatLogica is a deep tech enabling organisations to seamlessly unlock the full potential of modern hardware Advisors Clients / Leads Partners Investors

MatLogica Copyright 2021 MATLOGICA WHAT NEEDS TO BE DONE TO ACHIEVE PERFORMANCE 12 STEP 1 STEP 2 STEP 3 STEP 4 STEP 5 Drop-In Identify Record Execute Check Drop-in MatLogica library and replace ‘double’ to ‘ i double’ Identify and mark Inputs and Outputs for the target function Separate a single sample of target function for kernel recording Replace your normal workflow with flows to optimised kernels Run MatLogica diagnostics toolkit

The default solution for developers who want to extract a 100% performance from their hardware for computationally intensive software “The approach MatLogica takes to acceleration is novel in both its easy-to-use programming interface and high performance it achieves out-of-the-box. Straight-forward and minimal code changes, to make use of the libraries, offer leaps in performance for both xVA Pricing and Greeks Calculations.” “One notable feature of AADC is that complex model calibration procedures do not require any special attention as is often the case with other AAD approaches. For example, we “AADC-record” simple Dupire volatility calculations, regression-based continuation value calibration and even Monte-Carlo based Cheyette model calibrations, with no special care needed to back- propagate adjoints. This has allowed us to eliminate a large amount of complex and hard to maintain code.” "MatLogica offers a unique tool for developing machine learning applications in   C++. It combines the simplicity of object-oriented design and the performance of graph- based languages. For applications like time series analysis, recursive neural nets, and similar, the performance of MatLogica AADC is impressive.   Due to excellent vectorization and good support for multithreading, the execution on a CPU is very fast. In comparison with Tensorflow or Python, custom design implementation from scratch is simpler, making it an excellent tool for the implementation of novel algorithms. Using only C++ for implementation   improves portability and development of large software systems.” "Matlogica offers a graph-based framework in C++, optimized tor finance applications on modern hardware aware and complete win state-or-the-art AAD and proprietary on-the-fIy compilation. Quants will find that Matlogica seamlessly fits with their libraries and effortlessly accelerates pricing and risk by impressive amounts. And they will love the interaction with the founders" MatLogica's product changes the paradigm for quantitative software development eliminating the need to invest in optimisations. Quants can now focus on the models, and performance will be taken care of by MatLogica's JIT compiler "It's hard to develop well-performing models in C++.   I've been very impressed by performance gains up to two orders of magnitude obtained by MatLogica after some integration work on QuantLib, and by the fact that the same work also enabled AAD; especially considering that the library contains hundreds of thousands of lines of code developed over more than 20 years." Mahesh Bhat Principal Engineer, Intel Senior quant at Tier-2 bank ( case study available) Dr. Roland Olsson Associate Professor of Computer Science, Østfold University College Paul A. Bilokon CEO, Thalesians Ltd Visiting Professor, Imperial College London   Luigi Ballabio Co-founder and administrator of QuantLib Antoine   Savine Author of “Modern Computational Finance” * See a whitepaper led by Intel demonstrating up-to 1770x performance improvements for XVA pricing ** See Roland’s research demonstrating state-of-the-art neural network architectures for time series analysis based on MatLogica AADC

MatLogica Copyright 2022 MATLOGICA CASE STUDY - PROVEN RESULTS 14 5 min 18 sec* Manual AD AADC Standard Structured Note Performance NEW REVENUE STREAMS grid costs 50% LOWER Extended set of products via on-line channels 16x faster faster IT turnaround PRODUCTIVITY INCREASE 3-4x QUANT OUR CASE STUDY SHOWS HOW A MAJOR EUROPEAN TIER 2 BANK USED MATLOGICA TO ACHIEVE: ✓ Calcula x ons 15-20X faster ✓ Improved risk management ✓ Lower infrastructure costs ✓ New revenue streams ✓ Faster IT turnaround ✓ Increased customer sa x sfac x on

MatLogica Copyright 2022 MATLOGICA COMPETITIVE ANALYSIS 15 MatLogica’s proposition is unique as it is the only tool that can deliver both, Automatic Adjoint Differentiation (AAD) and speed up for ‘forward’ repetitive calculations such as Monte-Carlo simulations, historical analysis, VaR and others. The table below looks at competition within AAD space, while the chart on the right looks at a larger context. MatLogica’s AADC can be also used for Machine Learning applications, with no need to use APIs like TensorFlow. This means better usability and performance than traditional approaches.

MatLogica Copyright 2022 MATLOGICA GEEKY NOTES An X-Value Adjustment (XVA, xVA) is an umbrella term referring to a number of different “valuation adjustments” that banks must make when assessing the value of derivative contracts that they have entered into Modern compilers for object-oriented languages are not optimized for calculation-intensive tasks. Extensive use of abstraction and virtual functions allows developing an easy-to-read code but the performance penalty is high. Writing high-performing vectorized and multi-thread safe code is a tedious and time-consuming task, while the result is usually hard to maintain. MatLogica’s accelerator utilizes native CPU vectorization and multi-threading, delivering performance comparable to a GPU. For problems such as Monte-Carlo simulations, historical analysis, and "what if" scenarios, the speed can be increased by 6-100x, depending on the original performance. OUR PARTNERS: Tachyum is behind the world’s first Universal Processor, that unifies the functionality of a CPU, a GPU, and a TPU into a single processor that delivers industry-leading performance, cost, and power efficiency Antoine Savine is a French mathematician, academic and financial derivatives practitioner with Superfly Analytics at Danske Bank, winner of the In-House System of the Year 2015 Risk award and RiskMinds’ Excellence in Risk Management and Modelling 2019 award.

MatLogica Copyright 2022 MATLOGICA PUBLICATIONS • Case Study: How a Major European Bank Revolutionised Their Front-Office Risk Management Using MatLogica AADC – Read full version here or short version here • Automatic Synthesis of Neurons for Recurrent Neural Nets – Read here • Automatic Implicit Function Theorem – Read here • Adjoint Differentiation for Generic Matrix Functions – Read here • More Than a Thousand-fold Speedup for xVA Pricing Calculations with Intel® Xeon® Scalable Processors – Read here • A New Approach to Parallel Computing Using Automatic Differentiation: Getting Top Performance on Modern Multicore Systems – Read here • Open-Source Benchmark demonstrating performance leap for valuation and AAD risk calculations using AADC on Intel Scalable Xeon CPUs – Read here • Remarks on stochastic AAD and calibration of financial models – Read here • AAD: Breaking the Primal Barrier - how merging Code Transformation and Operator Overloading techniques leads to a major performance boost – Read here . • Run existing CUDA analytics on both GPU and CPU, with added benefit of AAD – Read here • Product Presentation - MatLogica's AADC - Read here https://www.youtube.com/@matlogica2002