2024 CMP Industrial Projects

Novo Nordisk Research Centre Oxford (NNRCO) is an innovative target discovery and translational research unit with a focus on identifying novel therapies for patients with cardiometabolic diseases (e.g. diabetes mellitus, obesity). To identify novel drug targets, we employ a variety of advanced computational biology techniques (graph data science and machine learning) on a myriad of data sources (genomics, transcriptomics, etc).

In particular, the Machine Intelligence group, has developed graph-based machine learning models to perform link prediction onto knowledge graphs (KGs) and has explored the positive impact that enriching a KG with task-specific information to explore the molecular landscape of complex pathologies. Genetic information provides direct evidence of the relevance of a drug-target candidate to a disease, however, the identifications come without a clear molecular context.

One of the main aspects to consider for the hypothesizing of the mode of action (MoA) of a drug-target candidate is its tissue of action (ToA). Gene candidates identified through genome wide association studies (GWAS) can be contextualized into the biology of a tissue by using various bioinformatic functional annotations, gene expression and epigenomic data.

There is an analytical framework to predict scores for the ToA from a GWAS meta-analysis of type 2 diabetes (T2D) [1] that the student will implement inhouse. There is also a published knowledge graph called GenomicKB [2], with all publicly available genomic information, including GWAS associations, epigenome, transcriptome and other bioinformatic annotations. The student will enrich the inhouse version of GenomicKB with the findings of the ToA scoring software and compare the KG link prediction results of the original versus the enriched KG to identify genes associated to T2D.

The student will spend 8 weeks exploring: 1) the value of predicting ToA for genetic findings in helping hypothesize MoA of potentially relevant inhouse candidates, and 2) the impact that enriching a genomic KG with task-specific information has on link prediction algorithms to identify genes related to T2D. To this end, the student will first implement inhouse a software that has scored the ToA of genetic hits for T2D, and then will work with KG modelling approaches to evaluate link prediction algorithms onto the original and the enriched KG.

The student will have the possibility to come to the Novo Nordisk office in London, where they can interact with the other three members of the Machine Intelligence Dept. that are based there. In particular, with Marc Boubnouvski, who will be the main London Contact. M. Lisandra Zepeda Mendoza is based in Oxford, where the student can also visit if wished, and Lisandra will also visit the London office at an ad hoc basis for more in depth discussions with the student.

For the duration of the internship, the student will also attend the Machine Intelligence Department meetings, which are held online, as the other team members are based in Seattle and Copenhagen.

The student can also chose to work from home if wished. However, personal communication at the London office would be preferred.

Our full-time summer internships are designed to enable current students in a technical field to translate their skills and expertise into an industrial setting. You will join us at our Head Office in Cambridge, UK, for 10 to 12 weeks, where you will have the opportunity to work alongside our team of talented software and hardware engineers, mathematicians, quantum information theorists, computational chemists and physicists – all experts in their fields.

Every intern will have a dedicated supervisor and will work on a project designed to make the best use of their background and skills whilst developing their knowledge of quantum computing. We will support all interns to try and produce a concrete output by the end of the internship such as a paper, product, or software tool.

What you will do:
- Develop, devise and research algorithms and software to enhance Riverlane’s capabilities, contributing to one or more projects that are core to Riverlane’s goals
- Discuss ideas with colleagues and communicate work in the form of presentations and reports
- Develop an understanding of quantum computers and their industrial applications

Requirements
- A current student studying for a Master’s degree (including the final year of an integrated undergraduate and Master’s degree)
- Proven ability in computational and/ or theoretical work
- Experience with at least one programming language
- Excellent critical thinking and problem-solving ability
- Strong communication skills, both written and verbal
- Ability to take initiative and to work well as part of a team
- An interest in quantum computing (extensive knowledge or experience is not required)

For more information, please visit https://www.riverlane.com/jobs/internships To apply please visit https://apply.workable.com/riverlane/j/A4807E07E1/ or email your CV and covering letter to jobs@riverlane.com by 21 January 2024.

Everyone is welcome at Riverlane. We are an equal opportunities employer and encourage applications from eligible and suitably qualified candidates regardless of age, disability, ethnicity, gender, gender reassignment, religion or belief, sexual orientation, marital or civil partnership status, or pregnancy and maternity/paternity.

Studies have shown that women tend to apply to jobs if they meet all or almost all of the requirements whereas men apply even if they meet only some of the requirements. If that sounds like you then please apply – we are happy to review your application and let you know if we think you might be a good fit.

If you need any adjustments made to the application or selection process so you can do your best, please let us know. We will be happy to help.

Important notes:
1. We are only able to consider applications from (i) current university students who are UK/ Irish nationals OR who have UK/ Irish settled/ pre-settled status OR Master's students who are on a Tier 4 visa.
2. Regrettably ware unable to provide letters of sponsorship or letters of support for sponsorship applications for your internship scheme. You must be available full-time for 10 to 12 weeks over the summer, preferably between late June and late September 2024. 3. We require a signed agreement from you assigning the ownership of any IP produced during your internship to Riverlane. 4. Internships are based at our Head Office in Cambridge, UK.

The Barclays FX electronic trading business streams tradable FX prices to institutional and corporate clients for them to transact fully electronically. This activity generates FX risk, which is also managed in an automated algorithmic fashion. There is on average a non zero cost to clearing that risk.

Optimal management of risk reduces cost and allows the streamed prices to be more competitive. From a research perspective this touches on portfolio management and execution strategies for trading in wholesale electronic markets.

We propose 3 distinct projects within this area. A potential intern can express interest in one or more of these.

Project 1: Optimal execution schedule in the presence of exogenous price prediction signals

Algorithms are often used when a participant wishes to buy or sell a given quantity of currency or security on public exchanges. These algorithms will split the total amount up into a schedule of quantity to be executed over a period of time. The general optimal execution schedule is a well studied problem. When faced with the problem of optimal execution scheduling, a market participant would typically balance a variety of factors including market impact and risk aversion. The aim of this project is to derive a theoretical extension of this framework in the presence of exogenous price prediction signals, and apply it to automated risk management of FX positions. In particular considering the case when we have both size and timescale for the price movements, or curve thereof.

Project 2: Identification of persistence of iceberg liquidity levels in external markets

In lit orderbook markets participants can place iceberg orders at a given price level. The total amount of these orders is not visible in the market data stack. It is possible to accurately estimate whether a particular level in the book has iceberg liquidity following a market trade. The aim of the project is to study the persistence of such iceberg price levels in time. This information can be used both in market making and execution logic.

Project 3: Optimal management of multiple portfolios in the presence of portfolio specific meta data

In the course of FX market making to clients, an electronic business will have a resultant position to risk manage. The composition of that position is heterogenous, and as such our risk aversion or direction view may depend upon the meta data of the underlying breakdown of that position. The purpose of this study is to establish optimal strategies for management of such a portfolio of risk, with a view to minimisation of cost of hedging. Common to all projects: The intern will have access to our historical data set of market data and trading activity. They will also have access to our python dev and compute environment for conducting this research. The internship will be at Barclay’s head office in London, and they will be working closely with the team of quants responsible for the development of trading algorithms.

We would appreciate if you could submit your formal application via the following link:
https://search.jobs.barclays/job/london/electronic-trading-associate-off-cycle-internship-2024-london/13015/54029345952

The application entails submitting your CV and some standard testing required for all Barclays applicants.

Please note the following deadlines:

• Applications (CV submission, candidate details, online assessments) must be completed via the above link by the deadline of 5 February 2024.
• Successful candidates will receive, as part of the recruitment process, an online assessment to showcase their technical skills by solving job-related problems. Deadline to complete the assessment is 13 February 2024.
• Successful candidates will receive an invitation to take part in person in an Assessment Centre at Barclays, which is planned to be held in late February.

MM Flowers is a cut flower importer and distributor based in the UK and Europe. MM was established in 2007 to provide the leading UK retailers (and laterally European retailers) with a transparent, sustainable and innovative interface between the grower & the retailer. As part of a vertically integrated supply chain, MM is owned by Elite, the largest flower grower in the world, based in Colombia, Ecuador & Kenya; VP, the largest rose grower in East Africa, based in Kenya & Ethiopia; and AM Fresh, a leading supplier of fresh fruit to major retailers in the UK and Europe. Collectively, the group is involved in all aspects of the supply chain from breeding, growing, freight, marketing, and distribution. Also part of the wider group, APEX Horticulture is an independent, professional research and development business, offering bespoke services for cut flowers and plants. APEX is based in a purpose-built testing centre adjacent to MM Flowers, an optimal position in the chain ensuring APEX can deliver high quality research on the true performance of flowers and plants subjected to actual supply chain conditions.

Whilst cut flowers are grown worldwide, many of those imported and sold in the UK / Europe are from Equatorial regions. Typically these flowers are packed dry in boxes and shipped to the UK / Europe by plane, although due to various factors, including the environmental impact and rising freight costs, alternative solutions to air are being explored. One of those may be transitioning the transportation of cut flowers from air to sea freight which has many benefits, but substantially increases the freight time (a 4-5 fold increase from South America to the UK, and potential a 10+ fold increase from Kenya to the UK). Whilst there are some flower types that have been successfully transitioned to sea freight, historically attempts with roses, the most important cut flower, have been unsuccessful. However, given the recent collective global focus on reducing harmful environmental practices, for the last 18 months, all of the stakeholders described above have been carrying out a series of projects to understand the impact of transporting roses (amongst other flower types) from both Kenya & Colombia to the UK & Europe by sea freight.

The project work described previously has largely been conducted by APEX to understand firstly the potential of different cultivars to be transported by sea, and where there were visible impacts on either the performance or quality of the flowers, what factors / processes may be influencing this. However, commercial trials have also been undertaken on select cultivars to determine what impact the transition to sea freight may have on both the practicalities of importing and distributing flowers, and most importantly, the final consumer experience. A large amount of data is therefore available from both the project work carried out, but also initial commercial data, supported by an even more extensive historical datasets on roses transported to the UK / Europe by traditional means (air freight). Whilst a huge amount of insight has already been gained from this information, due to the sheer number of factors potentially involved (e.g. cultivar, grower, farm location, growing conditions, freight time, etc), it is believed that this can be exploited further to inform strategic decisions going forward. As such, there are a number of areas that MM Flowers, APEX and the wider group would like to explore further –

• Based on the information available, are there key factors that determine the success / failure of transporting roses by sea freight?
• Do specific growers / farms produce roses that are better suited to sea freight transportation?
• Can MM Flowers data (such as quality assurance inspections, and sales, waste & complaints) be used to determine the success of commercially transporting select rose cultivars to the UK?

Silvaco is the software engineering company developing the tools to assist in manufacturing of semiconductor devices. In UK office we mostly work on 'process simulation' side -- mathematical modelling of the processes used in manufacturing.

One of such processes is implantation -- bombardment of piece of (typically) Si with ions (dopants), to change the electrical properties of the target in specific areas. To predict the final ion distribution we use Monte Carlo simulation -- follow the path of large number of ions, as they fly through the structure. There are several effects involved, one of them -- ions 'bouncing' off the crystal grid atoms. The collision and its effects are described by mathematical models. These models are currently approximated through the explicit empirical expressions. We are interested in testing how accurate these explicit formulas actually are and how replacing them with more sophisticated approaches may affect the simulation results.

Your task would be to review the literature and to suggest and to implement the required algorithms.

In this project, the student will explore how we can use diffusion models (DMs) to predict splicing and splice-sites. DMs are a generative AI method that uses insights from stochastic processes, such as Markov chains, to diffuse data (i.e. add noise to data until it follows a vanilla distribution, such as a Gaussian or uniform distribution). The models then learn how to "un-diffuse" the vanilla distribution to generate an unseen data point that follows the data distribution closely. DMs have been successfully used to make predictions about proteins [1] (such as their structure) from sequence information only. The student will build a DM, similar to the ones applied in the protein space [2], and train it on splicing data, which can be modelled in a similar way. They will then benchmark their method against state-of-art splice site predictors (such as [3], which are not DMs) at making tissue-specific predictions of splice-sites. If time permits, we can also evaluate zero-shot abilities of DMs for predicting splicing by withholding certain tissue contexts during training and then evaluating predictions in unobserved tissues.

Applicants should have experience with coding in Python. Experience with the PyTorch library and with version control (e.g. through git) are a plus. Prior knowledge in biology or genetics is not a requirement.

The intern will have the possibility to come to the Novo Nordisk offices in London (near King's Cross) or Oxford. The intern can work remotely or in a hybrid setting (from the UK) if they wish, although we expect them to be present in-person for on-boarding (in the first week) and the hand-over of the project (in the last week) in person.

Throughout the internship the intern will attend team meetings of the Computational and Statistical Genomics team to get exposure to a broader range of research and the opportunity to interact with other scientists working on related projects.

In industrial contexts, a multitude of power infeed errors arises from challenges associated with grid impedance. Precise estimation of grid inductances or resistances is pivotal in preventing infeed failures, reducing downtime, and optimising control processes. The traditional technique for grid impedance estimation typically entails temporarily halting industrial drives and introducing specific signals during commissioning.

However, a pioneering alternative is presented through the application of machine learning/artificial intelligence (ML/AI) techniques. This innovative approach enables continuous real-time estimation of grid impedance with an impressive precision of within +/- 10% deviation. The implementation of this ML/AI-based methodology not only eliminates the need for downtime during the estimation process but also empowers industrial systems to dynamically adapt to changing grid conditions.

This advanced capability not only contributes to the seamless optimisation of the entire drive system but also acts as a pre-emptive measure against potential infeed failures. For a discerning mathematics student at Cambridge University, this cutting-edge approach offers a fascinating intersection of mathematical modelling, data analytics, and practical industrial applications, showcasing the transformative potential of ML/AI in mitigating complex challenges within the realm of power systems.

Accurate estimation of grid impedance by reduced features

Key Learning Outcomes:
 - Acquired proficiency in industrial drive systems and demonstrated expertise in working with MATLAB/Simulink simulation models.
 - Demonstrated familiarity with Python scripts developed in prior projects.

Project Outcomes: 
 - Attained a comprehensive understanding of the intricate relationship between various features and target variables.
 - Pursued accurate estimation methodologies without compromising system performance.
- Applied ML/AI methodologies to the current Siemens drive systems, focusing on G120C or S120 models.

As a value driven company, Siemens is constantly trying to offer the same functionality using the smallest amount of resource. Whilst trained Deep Neural Networks (DNNs) can be incredibly effective in solving many of the industrial automation / control tasks where our products are often deployed, the size and computational overhead required can become prohibitive.

We would like a student to explore DNN compression techniques which greatly decrease the number of Neurons required to implement a given solution, with an aim to realise a solution on a tightly restricted processor by using the packages provided by the Fraunhofer in the given link.

In addition to the compression, in the same way that ‘Divide and Conquer’ algorithms can drastically reduce the computational cost of the discreet Fourier transform; we would like the student to try to generalise the problem of Neural Network computation in such a way that the number of operations required to implement the inference part of the Neural Network can be reduced.

Short-term interest rate (STIR) options are derivative contracts which give the option holder (long party) the right to a payout based on the difference between the underlying rate on expiry of the option and a pre-set level for the rate (i.e. the strike price of the option). The right comes at a price, the premium for the option – which depends on the strike and the time to maturity of the option, inter alia. Equivalently, for a given strike and time to maturity, the option premium can be mapped to an (implied) volatility level, and consequently, to a volatility surface for the strike and time to maturity domains.

Short-term interest rate volatility opportunities increased with the volatility surface shifting regime since late21/early22. For example, taking a benchmark European short-term interest rate (Euribor or more recently Estr) option markets were typically pricing 1 to 2 basis point move per day pre-hiking cycle vs 7bps/day+ territory more recently. In low carry environment, fixed income investors can absorb a relatively small rise in realised volatility before hedging/reducing exposure (which, once triggered, can contribute to further hedging). The new regime makes it particularly interesting to explore those products valuation vs more liquid instruments.

The expectations of the project are:
- Fitting volatility surfaces under different assumptions
- Pricing Capfloors and options on futures in a consistent normal volatility framework
- Potentially exploring the relative valuation vs swaptions

No prior knowledge of the instruments is required but a basic understanding (or interest) in the Fixed Income market and products (futures, otc swaps) will definitely help.

We will provide sample data: premium and implied volatility of capfloors and future options + required interest rate curves.

These are the different tasks for the project that we can adapt to the student/context:
- Stripping Euribor/SOFR vol surface: Unlike swaptions where one option price = one implied vol, Euribor/SOFR vol surface need a stripping logic of the liquid instruments (capfloors and future options)
- Producing a cap/floor pricer through quantlib
- Translating the spread to the rest of the surface into forward volatility and/or correlation of short-term interest rates. This can be used to display risk or to study relative valuations (the project can take different directions from here).

There are several approaches to CPD, including online vs offline and Bayesian vs Frequentist vs non-parametric. The student should first conduct a literature review of the various methodologies and then investigate one or more specific methods in depth using financial data. Some examples are:

- Online CPD methods for measuring asset volatility and correlation
- Online CPD methods using high-frequency price data
- Offline CPD methods for analysing portfolio manager returns