Daniel Helo Puccini

giving awesome AI tools to students

-Made community across difference by engaging with residents in halls through active listening and team communication strategies

- Built an internal MCP-enabled AI Agent for log diagnosis and coding support with LangChain, TypeScript, and Azure AI Foundry. - Automated an end-to-end workflow with AI for incident management and PRs reducing task time from 3+ days to <30 minutes.

Currently working as an Amazon Robotics Co-Op. Automated Testing Engineer.

- Developing a ChatBot for K-12 tutors to enhance knowledge. Leveraged AutoGen to process input through a Multi-Agent LLM system. - Managed user dataset with custom MongoDB endpoints for the AutoGen group chats. Created front-end with React.

- Built a Computer Vision model through supervised learning to detect weather conditions and inform drivers of hazards as part of portfolio project. - Connected with mentors and stakeholders in the AI industry to understand trends and address issues in diversity, equity, and inclusion. - Learnt about Ethical & Constitutional AI through topics like algorithmic bias and model alignment to ensure technological and social prioritization of safety, equity, and inclusion for all in the AI industry.

(now working as advisor) - Democratizing AI knowledge by hosting interactive workshops for club members. - Founder of AI Ethics discussions to understand the interdisciplinary nature of AI. - Improving people's careers by bringing awareness to the importance of AI skills in the workplace.

- Previously led 12 development teams towards creation of full-stack AI products. Leveraged communication and DevOps strategies. - Garnered 100s of users across the product suite (music classifier, AI note-taking app, etc.) by deploying on Vercel with Flask and Next.js.

Facilitating impactful learning experiences by employing active listening and communication techniques to increase attendee engagement and learning.

- Led a development team to implement data augmentation techniques and incorporate Spanish knowledge to improve a morphological tagger. - Leveraged the Stanford Staza NLP library to increase accuracy by 30% for Latin American Spanish, achieving a new State of the Art. - Writing a technical paper to share methodology with the broader research community, with aims of increasing visibility of Spanish research

- Developed a research tool library in Python for integrating domain knowledge in deep learning architectures by prompting LLMs - Designed an evaluation framework with Gradio + Next.js that gathered over 50 responses in various metrics to improve the toolkit.