Prompt Engineer, Contractor

As a Prompt Engineer at Sesame, your primary responsibilities will include developing complex prompts for specific internal tasks, collaborating with our engineers and scientists to refine these prompts, and gathering and analyzing data to measure their effectiveness. This role requires creativity, technical skill, and the ability to work closely with a multifaceted team.

To be considered for the Prompt Engineer role at Sesame, you should be fluent in English and have familiarity with data annotation and scripting, ideally through coursework or relevant experiences. Preferred candidates will have experience with large language model prompting and excellent writing and communication skills, which can be valued through a degree in English or similar fields.

Yes, the Prompt Engineer role at Sesame is primarily remote, but candidates should be located near New York City, as occasional office visits are expected for team collaboration and meetings. This allows for a flexible work-life balance while still maintaining team cohesion.

The expected workload for a Prompt Engineer contractor at Sesame is between 10 to 15 hours per week. This part-time commitment allows flexibility for individuals while ensuring you can contribute meaningfully to the development of our innovative projects.

Sesame values a diverse and inclusive workplace culture where everyone feels respected, valued, and empowered. As a Prompt Engineer, you'll be part of a team that embraces diversity and actively seeks individuals from various backgrounds to contribute to our vision of creating lifelike computer interactions.

When developing prompts for large language models, it’s essential to start with a clear understanding of the task at hand. I research existing prompts and outcomes, identify gaps in the current methodology, and then create iterative prompt drafts. Testing each prompt with real data helps refine them further.

In one project, I automated data annotation for a text classification task using Python. I wrote scripts that processed raw data and labeled it based on specific criteria, significantly increasing accuracy and reducing the time needed to annotate manually.

I have experience using ChatGPT and Claude through personal projects where I experimented with generating conversational replies. I learned to craft effective prompts that yield specific responses, which gave me a comprehensive understanding of how to interact with these models productively.

In a recent internship, I collaborated with a team of engineers to develop a chatbot. We held regular meetings to discuss methodologies and results, shared feedback on each other’s prompts, and iterated the prompts based on comprehensive testing to improve user interaction.

To measure the effectiveness of prompts, I utilize ground truth data analysis and user feedback. By A/B testing different prompts and evaluating their performance against predefined metrics, I can gauge their success and make data-informed adjustments.

One of the main challenges I faced was understanding the biases in AI models. During a previous project, I had to ensure that the data I was using for prompting did not perpetuate those biases, which meant extensive research into the datasets and how they were formed.

I prioritize tasks by assessing deadlines and impact. I create a schedule to manage my time effectively, focusing on high-impact projects first. Regular check-ins with the team help keep everyone aligned with the project’s goals and timelines.

Effective techniques for writing prompts involve being specific but also allowing flexibility for the AI's responses. Using open-ended questions, providing contextual information, and iterating based on results are techniques I find lead to better outcomes.

I stay updated by following several AI-focused publications, participating in online forums, and attending webinars and conferences. Networking with other AI professionals also provides valuable insights into emerging trends and best practices.

After receiving feedback on a specific prompt set, I analyzed the responses and identified patterns in user interaction. Based on this analysis, I rephrased some prompts for clarity, which led to a significant increase in user satisfaction and engagement.