Harel Wilner

As a LLM Engineer I have been pivotal in advancing automated digital media strategies that relate to Israel-Gaza war. I worked with Hugging Face, OpenAI API, Llama Index technologies to create online research agents, integrated with Twitter API for data collection, analysis and building geolocation features.
● Optimized GPT-4 API expenses using knowledge graphs and working with AWS & GCP cloud environments.
● I enhanced the RLHF technique by using Google PaLM to implement - 'Optimization by Prompting', and utilize PFET & LoRA to fine-tune models effectively.

MatchUp was created to enable personalized recruitment processes tailored to both the candidate's and the employer's needs. As a researcher, I am developing tailored job description analyses using NLTK with PyTorch and Word2Vec, analyzing candidates’ resumes using LLM, and utilizing transfer learning and prompt techniques to calibrate LLM Transformers using both LangChain and Hugging Face modules.
● Saving time for the 'mentors' who edit and comment on the resumes.
● Reduced mismatches between employers and candidates by more than 80%.

Neuronics is a consulting company in the data science field, with an emphasis on computer vision, creating solutions to improve crop quality and manual labor tasks. I was responsible for implementing and training advanced computer vision models, including YOLO, Faster R-CNN, U-Net, and Mask R-CNN, to identify different types of coffee beans. My role involved preprocessing and cleaning large datasets and optimizing model results to meet the strict requirements of our clients. Utilizing Python and TensorFlow, I contributed to creating impactful solutions that:
● Reduced future manual labor processing by more than 80%, saving hundreds of thousands of dollars.
● Achieved over 93% accuracy in detecting coffee bean types, capturing the slightest details.