Hi, I am Aastik
Senior Machine learning engineer, at Samsung Research and a vagabond
Try my jingaxAI to know me better!
Hi there!
I'm Aastik, an ML Engineer at Samsung Research. With a Master's in Technology from IIT Bombay (2023), my journey in the tech world has been nothing short of exhilarating. My thesis focu into the fascinating world of ML in constrained setups, and now, I’m navigating the cutting-edge realm of Vision Language Models.
I’m a quick learner and adapt like a chameleon! When I’m not immersed in algorithms and data, you can find me scoring goals on the football field, hitting sixes in cricket, making a splash while swimming, or serenading with my flute. Life's a blend of code and rhythm for me!
Let's connect and create something extraordinary!
Work Experience
Download pdf? Click here
Samsung Research India
2023-Present
As a Senior ML Engineer, I transitioned flagship devices from tag-based to natural language search using Vision Language Models (VLM), improving search relevance for over 10 million devices. I developed NEXIN (submitted in EMNLP 2024) , a preprocessing technique for exclusion search queries, boosting accuracy by 85% and securing a patent for my work.
Masters Thesis - IIT Bombay
2021-2023
Published a paper titled “Framework for Co-distillation Driven Federated Learning to Address Class Imbalance in Healthcare” in CODS-COMAD 2024. Worked on machine learning in federated settings, focusing on optimization through subset selection and knowledge distillation. Led a joint team from IIT Bombay and MIT in advancing federated learning research.
Fullstack Developer - IdealVillage, Bangalore
2020
Developed key features for the platform, including a complete wallet system for transactions, payment API integration (PayTM, PayUMoney), and a subscription system. Implemented an affiliate dashboard with analytics tools like charts and maps, enhanced the payments page design, and expanded the affiliate system to support multiple user types with varying permissions.
Bachelors Thesis - JSS Academy
2017-2021
Led a team of four to develop a platform for identifying collusive users on Twitter, achieving 87% accuracy. Built web automation tools to collect over 1,500 accounts, compiled datasets using the Twitter API, and proposed 65 attributes for bot detection. Created a Django-based platform for real-time analysis and introduced confidence indexes to assess trustworthiness. Published the paper “Machine Learning-Based Identification of Collusive Users in Twitter Stream” at ComPE 2021, winning the Best Paper Award.
Skills
Frameworks
TensorFlow, PyTorch, Keras, Pandas, NumPy, Scikit-learn, ONNX, PyTorch Lightning, MLOps, Django, Codeigniter, Android
Contact
Reach out to me on my socials or on my mail