Machine learning specialist with over 5 years of experience, I possess deep knowledge in MLOps, data science, mathematics, and algorithms. Whether integrating out-of-the-box products like ChatGPT, utilizing pre-trained models, or architecting custom solutions, I bring a wealth of experience to guide you towards the best solution for your problem - I don’t rely on copy-pasted repository hoping it will work for every problem.
I am adept at reading and implementing research papers, allowing me to recognize when client problems necessitate solutions that can only be effectively addressed through the application of methodologies outlined in specific papers. I spend my free time in Kaggle to keep myself abreast with implementation the latest techniques and study how rockstar data scientists make their solution highly efficient and robust
I am a firm believer in documentation. I invest time in commenting and documenting, recognizing the invaluable time saved during future revisions, especially in large-scale projects. I adhere to a systematic, scalable, and commonly accepted workflow. For instance, I prioritize project versioning with Git and favor containerized deployment with Docker.
With a track record of solving difficult algorithmic problems, I find joy in overcoming challenges. Clients have attested to how my assistance has rescued their projects.
Experience
-----------------------------
- Title recommendation system: Generate titles based on description, tags, and category using LangChain.
- Product matching from two retailers: This project involves a complex pipeline of data processing (including tokenization, translation, and Faiss indexing) and feature engineering (identifying frequent words, high-importance words, and price per unit). I utilize XGBoost to train a match presence detection model followed by regression for finding the closest match.
- Impact study of the adoption of smart homes on the UK grid: Conduct Monte Carlo simulations of appliance usage and train a demand forecast model using an ensemble of LGBM and CatBoost. Model tuning involves adjusting the lookback horizon of demand and the lookforward horizon of forecasted data (including weather, energy price, etc.).
- NxNxN Rubik’s cube solver using Iterative Deepening A* algorithm (IDA*): Employ heuristics to prune possible paths.
- Adaptive Neuro-Fuzzy Inference Systems (ANFIS) for ECG classification: Feature extraction involves digital signal processing in the time and frequency domain, waveform morphology, and statistical values. The ANFIS classifier is then trained with these features.
- Dashboard for weighbridge using D3: Develop a desktop GUI using Electron and utilize the D3 library for data visualization.
- Report generation for a hair product factory using Tableau: Integrate live data feeds queried with SQL and automate report generation.
Tech Skills
-----------------------------
ML Framework 🤖
- Scikit-learn
- Pytorch
- Tensorflow
- XGBoost
- NVIDIA TensorRT
Database Management 🗃️
- MSSQL
- MySQL
- PostgreSQL
- Firebase
- Pinecone
Cloud Technologies ☁️
- AWS (SageMaker, Lambda, EC2)
- GCP (Cloud Functions, Cloud Run, GKE, Compute Engine)
- Vast.ai
- Vercel
- Colab, Kaggle
DevOps, Automation, Server ⚙️
- Docker
- Kubernetes
- Ngrok, Nginx
- Linux
Web & API Development 🌐
- React.js
- Next.js
- Django
- Node.js
- Express.js
- Flask, FastAPI
JavaScript
