Dharmendra Thakur Email & Phone Number

Machine Learning Software Engineer with 2 years of experience driving impactful projects at Samsung Research Institute, Bangalore. I have a strong background in Multimodal Large Language Models (MLLMs), Large Language Models (LLMs), and Natural Language Processing (NLP) methods, alongside proficiency in Machine Learning and Deep Learning algorithms.I have contributed to key projects such as Smart Capture Solutions, Multimodal Large Language Models, and the SCREEN AI Project. Skilled in Python, Kotlin, C++, Data Structures & Algorithms, and Android Studio, I specialize in fine-tuning language models, resolving latency issues, and enhancing model accuracy.I am eager to leverage my expertise in Deep Learning and NLP to excel as a Machine Learning Engineer in a forward-thinking organization.

• 

  Machine Learning Engineer

  Samsung R&D Institute India - Bangalore

  Jul 2022 - Sep 2024 - Bengaluru, Karnataka, India

  Part of the Advanced Research Group at Samsung Research working on Multimodal, LLM and ScreenAI Projects. Research and Innovations• Submitted a paper titled ‘SeLoRA: Selective Low Rank Adaptation for Effective Task Expansion’ to the ICASSP-25 conference, experimented with Low Rank Adaptation methods, Patent idea submitted and currently in progress.Screen AI Solution• Developed Synthetic Data generation by using Mistral and Gemma, Data cleaning, and Data processing… Show more Part of the Advanced Research Group at Samsung Research working on Multimodal, LLM and ScreenAI Projects. Research and Innovations• Submitted a paper titled ‘SeLoRA: Selective Low Rank Adaptation for Effective Task Expansion’ to the ICASSP-25 conference, experimented with Low Rank Adaptation methods, Patent idea submitted and currently in progress.Screen AI Solution• Developed Synthetic Data generation by using Mistral and Gemma, Data cleaning, and Data processing scripts.• Developed screen layout-generalized pre-processor script for chat extraction from screenshots during inference.• Commercialization of this project is underway for the upcoming S25 Ultra series.Multimodal Large Language Model (MLLM)• Trained LoRA-configured MLLMs for VQA, Smart Capture tasks, Contributed to on-device and PC integration.• Optimized datasets, improved accuracy by 8% on SC, implemented UI-level enhancements for better performance.• Successfully demonstrated the multimodal large language model (MLLM) on device demo at the Techglanz event.• Optimized the training pipeline, achieving a 50% latency reduction, demonstrated on both device and PC demos.Smart Capture Solution• Contributed to a commercialized project that extracted screen information from screenshots using in-house OCR and rule-based approaches, and developed an Android app for user trials. Developed rules for info-extraction.• Dataset correction and rules optimization enhanced accuracy by 10%, achieved 90%+ accuracy for that modulePersonal Knowledge Graph• Improved app rendering latency by 60%, implemented user feedback modules, enhanced android app interface.• Received spot awards for contributions to the PKG project. Improved app performance, overall app design. Show less