AI and Software Developer · Technical Writer · Open-Source Contributor
I'm passionate about pioneering advancements in AI, with extensive experience in machine learning, deep learning, and generative AI systems. My journey spans from implementing traditional ML models to designing sophisticated AI workflows for automation and generative tasks.
Currently, I'm deepening my expertise in generative modelling by exploring diffusion models, flow matching, energy-based models, and generative world models driven by a strong interest in understanding how machines can learn and synthesize complex data distributions.
- Generative Modelling: Designing probabilistic models that capture complex data distributions through energy functions, diffusion processes, and flow matching — with a focus on scalable training and high-fidelity synthesis across continuous data modalities
- Multimodal Models: Developing multimodal models to integrate visual perception, natural language understanding, and other modalities for complex reasoning tasks
- Joint-Embedding Predictive Architecture (JEPA): Exploring self-supervised learning frameworks that learn world models through predictive coding in abstract representation spaces
- Few-Shot Learning: Creating adaptive systems that can generalize from minimal training examples using meta-learning and transfer learning techniques
- AI Agents & Automation: Building autonomous intelligent systems capable of complex decision-making and real-world problem solving through reinforcement learning and planning algorithms
| Project | Description | Stat |
|---|---|---|
| DeltaFlow | PyTorch library for flow matching and anatomy-invariant guidance alignment for radiograph generative pretraining |  |
| TorchEBM | PyTorch library for diffusion, flow-matching, and energy-based models, covering samplers and training utilities |  |
🤖 Generative AI and Agent Systems
| Project | Description | Tech Stack |
|---|---|---|
| PickSmart | Multi-agent AI assistant with agentic workflow | LangChain, LangGraph, MongoDB, Groq, FastAPI, Docker |
| DressMe.AI | Multimodal agent with image generation and feedback loop | VLMs, Diffusion Models, Feedback Mechanism |
| ai-teacher | Real-time voice-based AI avatar | Anam, ElevenLabs |
| SEAD-agent | AI research assistant with tool calling and RAG | VLMs, Function Calling, RAG |
| vlm-ollama | VLM inference on local deployment | VLMs, Ollama |
| SageMaker LLM Deployment | LLM deployment on AWS | AWS SageMaker, LLMOps |
🔬 Model Exploration
| Project | Description | Model / Approach |
|---|---|---|
| Generative Modelling | Implementations across generative paradigms | Diffusion, Flow Matching, Energy-Based Models |
| VLM GRPO | Post-training vision-language models | GRPO |
| VLM LoRA | Parameter-efficient fine-tuning of VLMs | LoRA, PEFT |
| Transformer Machine Translation | Neural machine translation built from scratch | Transformer, Attention Mechanisms |
| Post-Training Quantization | 8-bit quantization for LLMs | Symmetric and Asymmetric Quantization, BitsAndBytes |
🧠 Core ML / DL Projects
| Project | Description | Model / Approach |
|---|---|---|
| SeGAN | Financial forecasting using GANs | RNN, GANs, Time Series |
| DenseNet201 Classifier | Satellite image classification | CNN, Transfer Learning |
| GradCAM | X-ray image localisation | Computer Vision, Explainable AI |
| Multi-Regressor MLP | 2D landing control system | Neural Networks, Control Systems |
| MIMO Regressor | Biochemical production optimisation | Random Forest, XGBoost, SVM |
Master of Science in Artificial Intelligence (Distinction)
University of Essex
Bachelor of Science in Computer Science
Mahidol University
|
|
|
