LinkedIn | Resume | Website | Hugging Face
🚀 Machine Learning Engineer · AI Research Enthusiast · Generative AI Developer · Data Science Practitioner · Full-Stack Explorer
I build data-driven and intelligent applications that blend machine learning with modern web technologies.
From LLM-powered reasoning systems to interactive AI tools using React and Next.js, I focus on creating end-to-end products that are both scalable and impactful.
Curious and hands-on, I enjoy experimenting with new frameworks, optimizing workflows, and turning innovative ideas into working solutions.
- 🤖 Machine Learning & AI: Tree-based models (XGBoost, LightGBM), Time Series Forecasting, Model Optimization
- 🧩 Generative AI: Diffusion Models (Stable Diffusion, Flux), Style Transfer, IPAdapter, Inpainting, LoRA fine-tuning
- 🗣️ LLMs & Agents: Prompt Engineering, RAG, Multi-Agent Systems, LangChain, Vertex AI
- ⚙️ MLOps & APIs: FastAPI, Model Deployment (Azure / GCP), Automation Pipelines
- 🧍♂️ Computer Vision: Pose detection, Outfit transformation, Image segmentation (Grounded SAM, YOLO)
Languages:
Python · JavaScript · TypeScript · SQL · Bash
Libraries & Frameworks:
PyTorch · TensorFlow · scikit-learn · HuggingFace · LangChain · OpenCV · NumPy · Pandas · Matplotlib
Infra & Deployment:
AWS SageMaker · Azure · Google Cloud (Vertex AI) · Docker · FastAPI · Flask · Git · CI/CD · REST APIs
Front-End & Visualization:
React.js · Next.js · TailwindCSS · Plotly · Streamlit
Specialized Domains:
RAG Pipelines · LLMOps · Vector Databases (FAISS) · NVIDIA NIM · Multi-Modal AI · Diffusion Models · ComfyUI · Automation · Computer Vision · NLP
| Project | Description |
|---|---|
| EchoSeek | Developed an AI-powered search and recommendation system combining Llama-3 reasoning with vision-language embeddings for contextual product retrieval. Integrated LangChain, FastAPI, and Next.js for an end-to-end multi-modal RAG pipeline deployed on AWS SageMaker. |
| Order Amount Prediction | Built an ensemble regression model (XGBoost, LightGBM) for accurate sales forecasting, improving prediction stability and reducing variance across time periods. |
| Medical NLP | Designed a domain-adapted NLP pipeline using transformer-based models for extracting and classifying key entities from medical texts. |
| E-Commerce Website | Developed a full-stack web platform for product management and analytics, integrating real-time data visualization and recommendation features. |
- Building agentic AI workflows that reason, retrieve, and act autonomously
- Exploring multimodal systems that connect text, vision, and voice
- Writing cleaner, modular ML code for scalable AI pipelines
| Title | Description |
|---|---|
| Investigating Deep Learning Architectures for Robust Speech Emotion Recognition | Compared LSTM, CNN, and CLSTM models for speech emotion recognition, analyzing accuracy and robustness for emotion-aware HCI systems. |
| Leveraging BERT-Enhanced MLP Classifier for Automated Stress Detection in Social Media Articles | Built a BERT-Enhanced MLP Classifier achieving 95% accuracy on Reddit & Twitter data for stress detection and mental health analysis. |
| Multilingual SMS Spam Detection using BERT and LSTM | Conducted multilingual spam detection (English, French, German) using Kaggle data; Naive Bayes model achieved 98.1% accuracy. |
⭐ "Building systems that learn, imagine, and create."