Skip to content
View AsswinCR's full-sized avatar
:electron:
...
:electron:
...
29 followers · 59 following

Achievements

Achievement: Pull Shark

Achievements

Achievement: Pull Shark

Highlights

  • Pro

Organizations

@EddieHubCommunity @Design-and-Code @Blulearn-Hackathon @ML-Ninjas

Block or report AsswinCR

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Please don't include any personal information such as legal names or email addresses. Maximum 100 characters, markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
AsswinCR/README.md

Asswin Chithra Ramesh

AI Research | Computer Vision | Deep Learning | Generative AI

MSc in Artificial Intelligence @ Nanyang Technological University, Singapore

LinkedIn Gmail Google Scholar ResearchGate dev.to CodeSandbox

Python PyTorch TensorFlow C++ AWS Docker Kubernetes Hugging Face LangChain AutoGen LangGraph Crew AI LlamaIndex Weights & Biases OpenAI Ollama Anthropic

👨‍💻 About Me

AI Researcher with expertise in Generative AI, Computer Vision, Deep Learning, and Reinforcement Learning. I develop efficient, real-time AI models for edge devices, build scalable ML pipelines, and create LLM-based automation frameworks. Published Q1 research papers demonstrating technical depth and innovation.

With a diverse technical background, I:

  • 🔬 Develop and optimize neural network architectures (CNNs, 3D CNNs, UNet) for real-world applications
  • 🤖 Create scalable machine learning pipelines leveraging cloud infrastructure (AWS SageMaker, EC2)
  • 🧠 Build multi-agent LLM systems and automation frameworks using cutting-edge AI technologies
  • 📊 Design state-of-the-art solutions for complex problems across various domains
  • 📱 Deploy models to edge devices while maintaining high performance and accuracy
  • 📝 Publish research in high-impact journals demonstrating innovative approaches to AI challenges

🚀 Featured Projects

Automated Research Paper Implementation using Multi-Agent LLM Framework

Developed a fully autonomous multi-agent AI pipeline (GPT-4o, Claude, Autogen) for automating research paper analysis and implementation without human intervention.

Evaluating LLM Performance on Browser Tasks (WebArena Project)

Designed experimental evaluations to assess LLM capabilities in automating complex browser tasks using SFT and QLoRA, benchmarking 812 browser-based tasks.

Semantic Segmentation with Domain Adaptation using CycleGAN

Utilized CycleGAN for synthetic-to-real domain adaptation, improving U-Net segmentation accuracy on real-world images from GTA5 to Cityscapes datasets.

Blind Face Super-Resolution for Enhanced Image Generation

Built CNN-based super-resolution models (MSRResNet-B20, EDSR) achieving high-quality facial reconstruction from degraded images with a PSNR of 26.64.

💼 Professional Experience

Computer Vision Researcher

Rapsodo Private Ltd, Singapore (Sep 2024 – Jan 2025)

  • Developed lightweight CNN model for real-time golf club speed estimation
  • Reduced inference error by 80% (from 5% to 1.08%)
  • Designed novel CNN attention architecture for golf ball spin parameters

AI Researcher

Robert Bosch, Singapore (May 2024 – Nov 2024)

  • Developed RL pipeline (PPO) for HVAC energy management optimization
  • Achieved 87% reduction in training time and 60.78% reduction in cost
  • Used Ray and PyTorch DDP for scalable training across multiple instances

Multimodal Machine Learning Intern

Dementia Research Centre, Singapore (Nov 2023 – Jan 2024)

  • Implemented deep learning models (3D CNN, UNet) for early detection of Alzheimer's disease from MRI images
  • Improved diagnostic accuracy using Medial Temporal Atrophy (MTA) scoring
  • Enhanced model performance by analyzing and preprocessing multimodal medical imaging data

Generative AI Research Assistant (AI/NLP)

Nanyang Technological University, Singapore (Oct 2023 – Feb 2025)

  • Developed an end-to-end video validation pipeline using NLP and speech-to-text (OpenAI Whisper API)
  • Generated, timestamped, and validated video transcripts, automating previously manual workflows
  • Built a monitoring and validation UI with web scraping for enhanced data acquisition

Robotic Automation Intern

Yaskawa Pvt. Ltd., India (May 2022 – Jul 2022)

  • Programmed and simulated arc welding and palletizing robots using MotoSim EG-VRC & pendant
  • Modeled and analyzed 3D robotics palletizing algorithm and mechanics for mixed carton sizes
  • Developed a C++ algorithm to effectively use robot's memory during data transfer

Machine Learning Intern

Technocolabs Softwares, India (Sept 2021 – Dec 2021)

  • Developed a ML algorithm to predict stock market price movements based on O/H/C/L/V data
  • Deployed the predictive model on Heroku server for real-time predictions
  • Utilized Python, Django, Flask, and Heroku for the end-to-end implementation

🧠 Technical Skills

AI & Deep Learning PyTorch, TensorFlow, Keras, CNN Architectures (UNet, MSRResNet-B20, EDSR), Attention Mechanisms, GANs, Image Enhancement, 3D CNN
LLM & NLP Transformers, GPT, RAG, Autogen, LangChain, LlamaIndex, LangGraph, Ollama, SpaCy, Lite LLM
ML Operations AWS EC2, AWS SageMaker, Docker, Kubernetes, CI/CD, Git, DevOps, Apache Spark, S3, Parquet
Programming Python (Advanced), C++, Java, MATLAB, R, SQL, HTML/CSS, Prompt Engineering

📝 Research Publications

Pediatric Pneumonia diagnosis using stacked ensemble learning on multimodal CNN architectures

Journal of Multimedia Tools and Applications (Scopus Q1 - 81%)

Developed an enhanced image preprocessing pipeline using CLAHE and stacked ensemble deep CNN architectures for pneumonia detection. Achieved 98.62% accuracy, 98.99% precision.

Transfer Learning Approach for pediatric pneumonia diagnosis using channel-attention CNN architectures

Journal of Computerized Medical Imaging and Graphics (Scopus Indexed)

Implemented deep CNN models enhanced with cross-channel attention (inspired by Squeeze and Excitation Networks), demonstrating generalizability across diverse lung-disease datasets.
profile views

Pinned Loading

  1. Pediatric-Pneumonia-Diagnosis-using-Channel-Attention-Deep-CNN-Architecture Pediatric-Pneumonia-Diagnosis-using-Channel-Attention-Deep-CNN-Architecture Public

    Jupyter Notebook

  2. EPOCH-ESTIMATION-USING-DMD EPOCH-ESTIMATION-USING-DMD Public

    MATLAB

  3. MRF-for-Image-segmentation-Denoising- MRF-for-Image-segmentation-Denoising- Public

    Jupyter Notebook

  4. Multi-Arm-Bandits-for-Advertisement-click-through-rate-optimization Multi-Arm-Bandits-for-Advertisement-click-through-rate-optimization Public

    Jupyter Notebook