Dylan - AI Product Manager

Dylan: Your virtual AI PM avatar. Drives brainstorming on products/tech specs, shapes and analyzes new feature requests for existing products.

Comment

Inspiration

Imagine having access to a Technical Product Manager (TPM) who possesses a complete understanding of your company's products' data, tech stack and can be available at a moment's notice—especially when your human TPMs are overwhelmed with other priorities. This was the driving idea behind Dylan, our AI-powered TPM. Designed to act as a virtual collaborator, Dylan is always ready to assist, leveraging deep insights from across your organization to provide informed suggestions and facilitate strategic discussions. This ensures that no opportunity for innovation is missed due to resource constraints, enhancing productivity and decision-making across the board.

What it does

Dylan acts as a virtual PM, assisting teams by automating several critical aspects of product management. Key functionalities include:

  1. Brainstorming New Product Ideas: Facilitates the generation of new product ideas and technical specifications.
  2. Feature Request / New Products: Streamlines collecting and prioritizing new feature requests for existing products.

How we built it

The development of Dylan was a multi-faceted process involving:

  • Prompt Tuning: Tuned Dylan impersonating a TPM with existing knowledge of products and Tech stack.
  • Avatar Interface: Creating a user-friendly, avatar-based interface to humanize interactions with AI and enhance user engagement.
  • Integration with Tools: Seamlessly connecting with Jira and other project management tools to fetch data and manage storyboards effectively.

Challenges we ran into

Throughout the development of Dylan, we encountered several challenges, including:

  1. Gemini 1.5 Pro Rate Limits: One of the main challenges was dealing with the rate limits imposed on Gemini 1.5 Pro, which restricted the number of requests that could be made daily.

  2. Lack of WebRTC/JS Knowledge: Ability to stream Avatar + Script with low latency to the UI

  3. Complexity in fine-tuning the Technical Jargon: Training Dylan to accurately comprehend and generate technical language was intricate.

Accomplishments that we're proud of

We take great pride in the creation of Dylan, a virtual assistant designed to assist business users and stakeholders in brainstorming ideas and performing administrative tasks on behalf of a project manager. By integrating various technologies and services such as Gemini AI, Google APIs, Avatar API, Streamlit, Jira, and Gmail, we were able to create a one-of-a-kind, engaging user experience.

What we learned

This project allowed us to learn and work with several exciting technologies:

  1. User Experience and PM knowledge: Interviewed real TPMs and PMs about pain points and other day-in-life things.

  2. Gemini AI: We utilized Gemini AI's language generation (Gemini 1.5 Pro for better conversation build).

  3. API Integrations: We utilized several Email APIs, Jira APIs & D-ID Avatar APIs to stream the Text--> Speech Avatar into UI

What's next for Dylan - AI Product Manager

We see good potential for Dylan to be integrated into the organizational structure and Video chat platforms like Google Meet.

  1. Fine-tune past reports and presentations into the base LLM to improve Dylan's TPM and PM skills. [Easier with Gemini tuning by examples]

  2. Making Dylan have a constant grasp on the resource availability and code base of a product with tech specifications.

  3. Avatar * WebRTC integration to stream it into the UI. Integrating to Google Meets.

  4. Latency fixes with API rate limit enhancement.

Built With

Share this project:

Updates