- Joined
- Aug 2026
- How they joined
- Directly
- Before Cognition
- Department for Energy Security and Net ZeroData Lead
- Status
- Current
- Joined Cognition (operator of Devin) as an Applied AI Engineer in August 2026 and announced it verbatim: "Super excited to share that I have joined Cognition as one of the first Applied AI Engineers in Europe! Cognition operates Devin AI, the first autonomous software engineer. Looking forward to being part of the biggest revolution in computing since the invention of software!"Aug 30, 2026
- Being "one of the first Applied AI Engineers in Europe" coincides with Cognition announcing a new London office and UK team growth of more than 200% in six months (Oct 2026).Oct 22, 2026
- Left The Alan Turing Institute in August 2026, posting verbatim: "This was my last week at The Alan Turing Institute, which has been my home for the past four years. I joined the Research Engineering Group in August 2022, straight out of my PhD, where I had recently been promoted to Senior Research Software Engineer." He added "Now it’s time for new challenges and new adventures. More on that soon."Aug 15, 2026
- A second GitHub account, dsalvadorjasin, carries the bio "Applied AI @ Cognition" and hosts "A production-grade implementation of an Investment Portfolio Management System created for testing LLM translation of real world legacy applications".2026
- Current role is Applied AI Engineer at Cognition.
- At the Turing Research Engineering Group (Aug 2022 - Aug 2026) he supported "research software for digital twins, psychotherapy research, and weather forecasting", specialising in efficient data-driven weather forecasting tools.Aug 15, 2026
- Seconded in 2025 as GHGI Data Lead at the Greenhouse Gas Inventory within the Department for Energy Security and Net Zero, a year-long part-time secondment improving data engineering workflows and cloud analytics platforms and advising on digital initiatives within the National Atmospheric Emissions Inventory (NAEI).Dec 2025
- Co-creates SVD-ROM, an open-source Python package for scalable Reduced Order Modeling of large dynamical systems via Singular Value Decomposition built on Dask, alongside Robert Vava; he posted that the project "is actively looking for contributors".Nov 5, 2025
- Published "Principal component analysis and K-means clustering of fuel-air mixing in gas turbine combustors" in Data-Centric Engineering (vol. 7, 2026, DOI 10.1017/dce.2026.10039) with Loughborough co-authors A. Duncan Walker and Jon F. Carrotte, his NCCAT PhD research group.Feb 2026
- Contributed parallelized bootstrap aggregation (bagging) for Dynamic Mode Decomposition to the open-source PyDMD package, with Dask handling parallelization; he is a commit author on the BOP-DMD user manual.Dec 2, 2025
- Completed Helsinki's Master's-level Big Data Platforms course (grade 5/5, Nov 2025), earned Palantir Foundry Aware certification (Mar 2026) and took Stanford Mining Massive Datasets with Turing colleagues (Jan 2026).Nov 12, 2025I’m happy to share that I’ve obtained a new certification, with grade 5/5: Big Data Platforms from University of Helsinki! This is a Master's level course covering distributed computing, cloud and database technologies, including batch processing frameworks such as MapReduce and Apache Spark. #BigData #PySpark #Cloud
- Spoke at FOSDEM 2025 (HPC, Big Data & Data Science devroom, 2 Feb 2025) on "Explainable forecasting from big weather data: rapid and sustainable solutions" while a Research Data Scientist at the Turing.Feb 2, 2025
Cognition operates Devin, the first autonomous software engineer.
Technical advisor at an early stage biotech startup advancing technologies for early detection of Chronic Kidney Disease (CKD).
Development of robust and maintainable scientific software at The Alan Turing Institute, the UK's National Institute for Data Science and Artificial Intelligence. I specialised in the development of efficient data-driven weather forecasting tools.
I was seconded to the Department for Energy Security and Net Zero (DESNZ), where I provided technical advice to the UK Government on the digital upgrade of the UK’s Green- house Gas Emissions Inventory, harnessing cloud technologies and big data analytics.
I provided teaching support in the Aeronautical and Automotive Engineering Department for multiple subjects: Fluid Mechanics, Thermodynamics and Computational Fluid Dynamics (at undergraduate and postgraduate levels)
Doctoral research at the National Centre for Combustion and Aerothermal Technology (NCCAT), sponsored by Rolls-Royce plc. Thesis: "Latent variable analysis of mixing and soot production in gas turbine combustors". I employed high-fidelity numerical simulations of reacting flows (LES), multivariate statistical methods and reduced order modeling (principal component analysis, PLS regression, clustering) to analyze fuel-air mixing and soot emissions in jet engine combustion chambers.
Heating technologies research assistant under the supervision of Dr Aurik Andreu. During this summer placement I helped develop a suitable novel method of surface emissivity measurement for hot forging applications, where accurate measurement of temperature is crucial for the quality of the product and the energy efficiency of the manufacturing process. Knowledge of the numerical value of emissivity as a function of temperature and wavelength is essential in radiation thermometry. I worked closely with my supervisor and engineers from the AFRC's industrial partners. I also performed an in-depth literature review of radiation thermometry and numerical simulations of the heat transfer mechanisms in an emissivity calibration furnace using Matlab and ANSYS.
I carried out a numerical study of power-law flows in planar sudden expansions, funded by the EPSRC, which helped me expand my knowledge of computational fluid dynamics and non-Newtonian fluid mechanics, introducing me to academic research for the first time. Power-law flows in sudden expansions are relevant in several industrial and medical applications, such as in the case of abdominal aortic aneurysm. Blood is a non-Newtonian, shear-thinning fluid which can be modelled as a power-law fluid. Supervisors: Dr Mónica S. N. Oliveira, Dr Konstantinos Zografos
Super excited to share that I have joined Cognition as one of the first Applied AI Engineers in Europe! Cognition operates Devin AI, the first autonomous software engineer. Looking forward to being part of the biggest revolution in computing since the invention of software! 🤖 #NewJob
This was my last week at The Alan Turing Institute, which has been my home for the past four years. I joined the Research Engineering Group in August 2022, straight out of my PhD, where I had recently been promoted to Senior Research Software Engineer. The opportunities for professional development that the Turing has given me are invaluable. I have supported the development of research software for digital twins, psychotherapy research, and weather forecasting. I even had the chance to advise the UK Government through my secondment at the Department for Energy Security and Net Zero. I can’t think of a better place to learn as much as I did in the past four years. None of that would have been possible without my amazing colleagues at Research Engineering | The Alan Turing Institute, who are some of the smartest and nicest people I have ever met. Now it’s time for new challenges and new adventures. More on that soon.
A few of us from Research Engineering | The Alan Turing Institute will be completing a course on Mining Massive Datasets from Stanford University over the next few months. In the world of AI for Physics (climate modeling, engineering, etc.), you often have to deal with huge datasets that don't fit in your laptop's memory. You may use a HPC cluster or the cloud, but sometimes there are tricks that you can apply that allow you to work locally (e.g. out-of-core computing, blocked algorithms, randomized linear algebra, etc.). This is the main idea behind SVD-ROM (https://lnkd.in/dKJxgXu6), the open-source package I am developing to perform linear model order reduction on massive, real-world datasets directly on your laptop. Looking forward to starting the course. May the learning continue. #BigData #AIforPhysics
I’m happy to share that I’ve obtained a new certification: Certified Palantir Foundry Aware Professional from Palantir Technologies!
Very excited to share that my article "Principal Component Analysis and K-Means Clustering of Fuel-Air Mixing in Gas Turbine Combustors" is now publicly available in the Data-Centric Engineering journal. This work is based on research I carried out at the National Centre for Combustion and Aerothermal Technology (NCCAT) during my PhD. Fuel-air mixing in jet engine combustion chambers is incredibly complex and governed by multi-physics processes. The ability to control fuel concentration at the local level is important for combustion efficiency and emissions performance. In this paper, we present a novel statistical method using classical unsupervised machine learning to better understand the influence of individual air streams on local fuel-air mixing by post-processing Large Eddy Simulations (LES) of industrially-relevant combustors. The paper is available here: https://lnkd.in/eNsSRMV7 A 30 min video recording summarizing the paper can be found here: https://lnkd.in/eCMyEq8b #Engineering #Combustion #Aerodynamics #MachineLearning #CFD
Great to be back at FOSDEM 2026, the world’s largest Open Source Software conference! #OpenSource
I'll be presenting the work Robert Vava and I have been doing on the open-source SVD-ROM Python package at the next #PyData London Meetup on Tuesday next week. SVD-ROM is the first solution of its kind that enables you to apply SVD-based #ReducedOrderModeling at scale on any hardware, even on your laptop, thanks to the use of dynamic task scheduling and out-of-core computing powered by #Dask. For instance, this enables extremely cheap probabilistic forecasts of complex spatio-temporal dynamical systems (e.g. weather, fluid dynamics) by fitting Bagging, Optimized Dynamic Mode Decomposition models on arrays much larger than memory. Looking forward to presenting our progress so far! #BigData #DynamicalSystems #Forecasting #TheAlanTuringInstitute PyData London Meetup: https://lnkd.in/eEmbcNCG SVD-ROM: https://lnkd.in/dKJxgXu6
With my latest contribution to the open source #PyDMD package, you can now perform parallelized bootstrap aggregation when fitting Dynamic Mode Decomposition models to massive datasets. #Dask handles the parallelization for you, so this is super easy to set up on any machine. This opens the door to fitting DMD models with uncertainty quantification and producing probabilistic forecasts on huge arrays. Check out the PR on GitHub: https://lnkd.in/dgtHN8Uf #BigData #HPC #OpenSource
I’m happy to share that I’ve obtained a new certification, with grade 5/5: Big Data Platforms from University of Helsinki! This is a Master's level course covering distributed computing, cloud and database technologies, including batch processing frameworks such as MapReduce and Apache Spark. #BigData #PySpark #Cloud
SVD-ROM, the open source Python package I am developing alongside Robert Vava, is actively looking for contributors! SVD-ROM enables the application of Reduced Order Modeling (ROM) on #BigData using the Singular Value Decomposition, and it is laptop, cloud and HPC friendly. It relies on #Dask for distributed, out-of-core computing, #Xarray for labelled multidimensional arrays, and data formats such as #Zarr, #NetCDF and #HDF5. While our current focus is on computationally efficient and explainable weather forecasting, SVD-ROM has so much more potential, and has applications in fluid dynamics, combustion, finance, neuroscience or chemometrics, to name a few. If this sounds interesting, please do reach out! Link for the SVD-ROM repo on GitHub: https://lnkd.in/dN_NQNWe A demo showing how I can train an Optimized Dynamic Mode Decomposition model on a 35 GB slice of the ERA5 dataset and produce an accurate sub-seasonal weather forecast (about 6 weeks ahead) in about 5 minutes, directly on my laptop: https://lnkd.in/d5KNWqdM
