Studying by doing is one of the simplest ways to grasp important expertise for turning into a machine studying engineer. As a substitute of simply specializing in easy classification and regression fashions.
On this weblog, we are going to concentrate on superior machine studying tasks that may affect your resume and appeal to recruiters and hiring managers. We are going to find out about laptop imaginative and prescient tasks, speech recognition, inventory value forecasting, fine-tuning Steady Diffusion and Llama 3, multi-step AI agent purposes, and reinforcement studying. Additionally, you will find out about various instruments and ideas to construct and optimize these tasks.
1. Automated Picture Captioning
Automated Picture Captioning is a captivating venture that mixes laptop imaginative and prescient and pure language processing. The purpose is to generate descriptive captions for pictures. This venture makes use of convolutional neural networks (CNNs) for picture function extraction and recurrent neural networks (RNNs) to generate captions. Implementing this venture will reveal your means to work with complicated neural community architectures and deal with multimodal knowledge.
2. Automated Speech Recognition
Automated Speech Recognition (ASR) methods convert spoken language into textual content. This venture could be notably spectacular if you happen to work with a much less widespread language. It’s by far the most well-liked venture I’ve ever labored on. You possibly can even see for your self by going to the hyperlink kingabzpro/wav2vec2-large-xls-r-300m-Urdu and checking the variety of downloads.
On this venture, you’ll study to course of each textual content and audio after which fine-tune the wav2vec2 mannequin within the language of your alternative. If you’re in search of the code supply and information, you could possibly try the kingabzpro/Urdu-ASR-SOTA DagsHub repository.
After fine-tuning the mannequin, it can save you it on Hugging Face after which construct a real-time ASR app to deploy on the Hugging Face area, as proven beneath.
3. Inventory Worth Forecasting
Inventory Worth Forecasting includes predicting the longer term costs of shares utilizing historic knowledge. This venture could be applied utilizing varied machine studying methods reminiscent of time sequence evaluation, regression fashions, and even deep studying fashions like LSTMs (Lengthy Brief-Time period Reminiscence networks). You possibly can even use what you discovered from this venture to construct your individual buying and selling bot by integrating the inventory trade API.
4. Effective-tuning Steady Diffusion XL
Steady Diffusion XL is a strong mannequin for producing high-quality pictures. Effective-tuning this mannequin utilizing methods like DreamBooth and LoRA (Low-Rank Adaptation) might help you create personalized picture era fashions. On this venture, I’ve fine-tuned the mannequin utilizing 5 of my pictures, and the outcomes are wonderful.
You possibly can fine-tune it on particular cartoon characters and design your individual comedian e book utilizing Generative AI. This venture will showcase your experience in working with state-of-the-art generative fashions and your means to customise and optimize them for particular duties.
5. Effective-Tuning Llama 3 and Utilizing It Regionally
The tutorial “Effective-Tuning Llama 3 and Utilizing It Regionally” covers the venture of fine-tuning the most recent top-of-the-line open-source mannequin, Llama 3, on a medical dataset. The purpose is to construct a chatbot the place customers can ask inquiries to an AI physician.
All through the tutorial, you’ll learn to course of the information, use LoRA methods, optimize the mannequin and reminiscence, speed up the mannequin utilizing GPUs, and use varied instruments for merging, changing, and quantizing the mannequin.
On the finish, you’ll obtain the fine-tuned quantized mannequin and use it regionally utilizing the Jan software. This venture is just not solely enjoyable, but additionally an awesome studying alternative by which you’ll acquire a deep understanding of the way to troubleshoot varied points associated to fine-tuning giant language fashions.
6. Construct a A number of-step AI Agent utilizing LangChain
Constructing a multi-step AI agent includes making a system that may carry out a sequence of duties autonomously. Utilizing frameworks like LangChain, you’ll be able to develop AI brokers that may deal with complicated workflows.
On this venture, you’ll create an AI software that takes a consumer’s question to go looking the net utilizing the Tavily API and in addition generates Python code to make use of the information. The appliance will then use Python REPL to execute the code and return the visualization requested by the consumer. Earlier than beginning the venture, you’ll study concerning the Cohere API and its varied options.
7. Constructing MLAgent for 2v2 Soccer Recreation
Reinforcement studying is a strong method for coaching brokers to make choices in complicated environments. Constructing an MLAgent for a 2v2 soccer recreation includes creating an atmosphere, defining rewards, and coaching brokers utilizing reinforcement studying algorithms. Hugging Face affords hands-on tutorials for such tasks as a part of the DeepRL course which you can take without cost. This venture will showcase your experience in reinforcement studying and recreation improvement and your means to create clever brokers that may study and adapt.
Conclusion
Engaged on these superior machine studying tasks will improve your technical expertise and make your resume stand out to recruiters and hiring managers. Every venture covers totally different features of machine studying, from laptop imaginative and prescient and pure language processing to reinforcement studying and generative fashions. By showcasing your means to deal with complicated tasks and various datasets, you’ll considerably improve your probabilities of touchdown a high-paying machine studying job.