In my role as team lead & blackened developer at Pragmiq Oy, I guided the team in developing a Cardamom Price Prediction and Image Detection System, achieving 6-month predictive accuracy within a 20-25% improvement margin. The team was split into frontend and backend groups, with frontend members handling app design in Figma and development in Flutter-Dart, while the backend team, including myself, built predictive algorithms and image detection features.Focusing on backend predictions, I implemented time series and machine learning models, such as ARIMA and SARIMA. After multiple trials, we pivoted to a hybrid approach, achieving sustained accuracy in short-term predictions by 6 months while ensuring updates with daily web-scraped data, keeping our predictions in line with market trends.I integrated design, development, and real-time updates through coordinated project phases. I delivered a fully functional app that met market analysis needs and consistently provided reliable pricing data for Cardamom stakeholders.

As Head of the Engineering Denary for our annual Magnovite fest, I transformed participation and sponsorship outcomes by reimagining event design, selection, and outreach. To make the event more impactful, I increased participating teams by over 130% (from 75 to 172), retaining only 10% of previous events and creating 90% brand-new activities, designed to draw a wider audience. To build a strong team, I introduced a thorough interview-based selection process for department heads, event leads, and volunteers, emphasizing skills and commitment.On the financial front, I led daily coordination across six departments and drove the sponsorship team to achieve a 140% increase in sponsorship compared to prior years. Recognizing the need for long-term sponsorship growth, I strategically rescheduled Magnovite to align with sponsor cycles, ensuring easier sponsor commitments and financial stability.Through proactive leadership and planning, I set Magnovite on a new trajectory of success, establishing it as a high-impact, financially sustainable event with fresh, dynamic programming and a solid structure that future teams can build on.

Studying Imaging Techniques and Deep Learning Applications: I explored different imaging methods and how deep learning could address DOT's complex, ill-posed inverse problems. This research allowed me to understand the role of inverse modeling in DOT and its significance in improving tumor detection.Implementing Deep Learning for Image Reconstruction: I applied convolutional neural networks (CNNs) and other architectures to enhance DOT image quality, overcoming noise and distortion challenges. I collaborated on creating datasets, conducting Monte Carlo simulations, and developing models for accurate optical property estimation, which is essential in visualizing internal tissue structures.Evaluating Performance: My work also involved analyzing the effectiveness of deep learning techniques compared to traditional methods, using metrics like Structural Similarity Index (SSIM), Mean Absolute Error (MAE), and Jaccard Index. This helped refine our approach and validate the potential of deep learning in biomedical image processing.