Debayan Bhattacharya’s Post

For the third time in the last academic year, our team of Master's students from Sirius University won hackathon. This time we came second in the «Semantic Document Classification» category and received 200,000 rubles for our team. Team members: Ivan Butakov, Andrei Donskoi, Timofei Shchudro, Ali Ramazanov, Artem Medvedev The task of the hackathon was to automatically determine the type of a document (such as a law, an order, etc.) based on its text in order to speed up the work of companies when validating applications. The technological solution of the project involved parsing external data, fine-tuning a large language model and building a lightweight model based on decision trees for the service part. For the presentation, the team prepared an online service based on Docker containers with React and FastAPI frameworks for the client and server sides. The teams' results were evaluated based on two metrics: the accuracy of predicting the document type and the evaluation of the project's defence in front of technical and industry experts.

“How does technical excellence impact your daily work in software engineering?” Ans := It is magical! I was just lucky to have been taught by professors who put a lot of focus on problem analysis, domain modeling and solution design before they would start to code, I was just lucky there. When my professional career began then again I was very lucky to have joined a team that always tried to analyse the problem first and then try to design not just a solution but a maintainable and optimized solution. And I was extremely lucky to have found "Domain-Driven Design" by Eric Evans at the very beginning of my career as well. Of course I wasn't able to understand its full potential and couldn't implement it the way it should have been but I could see the light in the tunnel. Overtime, this practice of analyzing problems, designing maintainable and optimized solutions and constant search to find better ways made me cross paths with people like Alistair Cockburn, Ivar Jacobson, Martin Fowler, Robert C. (Bob) Martin, Valentina (Cupać) Jemuović, Karl Wiegers and many more from whom I learnt things, amazing things, really amazing things that changed everything around me! I was not only able to code, but I was able to finally write "maintainable, readable and extensible" code. I was able to hand-over the code base to junior developers and they could just swim in it with an ease that was just amazing to witness! All those long talks that people find boring about use-cases vs user-stories, functional vs non-functional requirements, aggregate vs entity vs value objects, business process modeling to conceptual vs logical vs physical data models...the endless discussions and constant refactoring to find the meaning of "ports" to reach the ideal hexagonal architecture, the endless fierce discussions what's a micro-service and what's a service to if micro-front-ends are even possible...all of these activities that people find boring are a part of my DAILY professional tasks and that's how I was able to reach at a point where I love my code base now! Not saying I am an expert, not at all, I am still a student, a learner, I might have said something wrong as well in my answer but I aimed for the stars sitting on a rocket...at least I will see the moon and the journey won't end with a crash!

As part of the 2024 LLM Hackathon for Applications in Materials and Chemistry my colleagues Hampus Näsström Pepe Márquez Michael Götte and I worked on an exciting project to convert spoken instructions into Electronic Lab Notebook entries for NOMAD using the open Large Language Model Llama. Our project aims to facilitate the documentation of lab experiments using audio input and extracting the information into structured data entries. This solution provides a hands-free and efficient way to document lab experiments and write structured data entries into an Electronic Lab Notebook (ELN) based on the NOMAD schema. It is particularly useful for labs where manual documentation can be cumbersome or impractical because scientists might need both hands in the glovebox while experimenting. The use of a local instance of an LLM is very important, as these experiments protect the IP of the scientist. In this implementation, we used the Llama3:70b model served via Ollama protecting the privacy and still offering an efficient solution for the text processing and structuring (via function calling). You find the code on GitHub https://lnkd.in/eEeKkNQN Great experience and I learned a lot!

🚀 Hackathon Experience: Conquering Challenges with Innovation 🚀 Over the past few days, I had the opportunity to participate in an intense but rewarding BR41N.IO hackathon hosted by g.tec, where I explored the exciting world of brain-computer interfaces (BCI). As a solo participant, I took on the challenge of coding, signal processing, and data analysis for SSVEP detection entirely on my own. 📊 Key Highlights: 1. Developed preprocessing methods (bandpass, notch filtering) for EEG signals. 2. Applied advanced techniques such as PCA, Canonical Correlation Analysis (CCA), and Filterbank CCA (FBCCA) for feature extraction. 3. Ran multiple classification models (LDA, CCA, FBCCA) and compared their performance across different sessions. 4. Achieved promising results with CCA, outperforming other methods in most cases. This experience pushed me to the limit, both technically and mentally, but it was an incredible learning journey. It also reminded me how crucial it is to balance innovation with well-being. Despite feeling burnt out, the sense of accomplishment from bringing this project to life has been immensely rewarding. 🔗 Check out my GitHub repository for all the code and data: https://lnkd.in/eweNxPNz 🔗 Presentation slides: https://lnkd.in/equU4CiA Thank you to everyone who supported me through this process! #hackathon #braincomputerinterface #BCI #machinelearning #neurotech #innovation #eeg

🎙️  I had the privilege of attending electrifying session conducted by IEEE SB KMCT CE and IEEE SB KMCT CEW “CRACK HACKATHONS LIKE A PRO” The session was conducted in KMCT CE , bringing together professionals and experts. 🔍 During the webinar, we delved into essential topics how & where you'll uncover the secrets to mastering hackathons and dominating coding challenges. The speakers MOOSA FAYAZ ,Basith P and Muhammed Zafar M. M. Shared valuable insights.

"Excited to have participated in Srishti 2k23, a National Level Technical Symposium hosted by PSG College of Technology. It was a great experience to engage in insightful discussions and learn from fellow participants. Looking forward to applying these experiences to future challenges! #Srishti2k23 #PSGTech #TechSymposium #LifelongLearning"

Super excited to talk and discuss deep learning for science and engineering at AMLD!

Fancy being part of a Research Hackathon with a social transformation non-profit, dedicated to creating a radically weller, wiser world through practical action? We're here for the pragmatic utopians who believe that the actions we take and the work we do CAN create the change we want to see, and need, in the world. Check out the details here for this online (and in person if you're in France) hackathon - 10th & 11th May - https://lnkd.in/e3J2Kp-F Come and join us and dive into our research projects and share your own! #Research #Hackathon #wisdom #sharing

Time to celebrate 🚀 DFRNT® got admitted to the MINC incubator in Malmö this summer, and we're preparing for solid growth over the years ahead. We're launching the Twinfox Platform that helps you design, leverage and collaborate on high quality structured information. We bring the software engineering collaboration model to the evolution of datasets: with data history, branch editing and merging of work; all this together with advanced information modelling and linked data capabilities. It is coupled with a deductive AI engine (datalog) that gives 100% correct results according to the rules it works with. Think AI with facts and rules! No LLMs, no magic: Pure logic, Linked Data Graph, Great results. What's not to like? Join our webinar where we officially launch the Twinfox Platform by DFRNT that helps you sharpen your competitive edge using high-quality structured information. Topics include more affordable plans, new platform capabilities, and introduction of hosted community sessions ahead on how to leverage the open source TerminusDB to navigate digital complexity. Join us 𝗦𝗲𝗽𝘁𝗲𝗺𝗯𝗲𝗿 𝟭𝟴𝘁𝗵 𝗮𝘁 𝟭𝟳.𝟬𝟬 𝗖𝗘𝗧 to learn more! Follow DFRNT to join the journey!! Register through the link in the comments below.

🎉We are happy to share that Nikola Konstantinov, a faculty member at INSAIT, and Nikita Tsoy, a doctoral student in Machine Learning, presented their paper “Provable Mutual Benefits from Federated Learning in Privacy-Sensitive Domains'' at AISTATS 2024 in Valencia. The paper is joint work with INSAIT summer interns Anna Mihalkova, ETH Zürich, and Teodora Todorova. 🔍The paper investigates the accuracy-privacy trade-off in federated learning, a central challenge as an uncompromising level of common privacy defenses often diminishes the possible benefits from collaboration. 🌐 AISTATS is a leading conference in AI and its links to machine learning and statistics. This is Bulgaria’s first paper in AISTATS in the nearly 27-year history of the conference.