Excelerate Prompt Engineering Internship
Team Lead & Evaluator — RCTFC Prompt Engineering Cohort
Led a 9-person cohort through a 4-week prompt engineering program, personally evaluating every contributor output against a 50-point rubric and feeding insights back to the team.
360° Evaluation
Three perspectives · averaged- 100%
- Self
- 100%
- Peer
- 92%
- Managerial
- 97%
- Overall average
9
Person team led
15
Outputs evaluated solo
50
Point quality rubric
4
Week program
Problem & Context
Excelerate's Prompt Engineering internship challenged participants to design, refine, and evaluate AI prompts under real-world quality constraints. As Team Lead, I was responsible not only for my own outputs but for coordinating a 9-person cohort and ensuring every submission met a rigorous 50-point standard.
My role combined leadership with quality assurance: I evaluated all 15 outputs from 5 contributors, identified recurring failure modes, and delivered structured feedback that helped the team improve week over week.
Framework & Methods
- Evaluation Rubric: A 50-point scoring system across seven dimensions including Clarity, Relevance, Hallucination Risk, Bias Risk, Creativity, Specificity, and Output Structure.
- Solo Evaluation: Personally scored all 15 outputs from 5 contributors to maintain consistent judgment and avoid evaluator drift.
- Feedback Loop: Consolidated common errors into cohort-level guidance, turning individual deductions into team-wide learning.
- Final Delivery: Produced and delivered the Week 4 individual final presentation deck summarizing approach, findings, and quality insights.
Key Findings
Hallucination and Bias Risk are the hardest dimensions to control
Across all 15 evaluated outputs, prompts that lacked explicit grounding instructions produced the highest variance in factual consistency and unintended bias — even when the final answer sounded confident.
Structure beats creativity when consistency matters
Contributors who used role framing, step constraints, and output schemas produced more repeatable results than those relying on open-ended instructions. Evaluability improved sharply with explicit formatting.
Team quality scales with feedback loops
As the sole evaluator, I consolidated recurring failure modes into a shared feedback loop. This lifted the cohort's average output quality by the final week and reduced repeated rubric deductions.
Recognition
This work earned a Star Performer award for the Excelerate Prompt Engineering internship, recognizing both the quality of my individual outputs and the accountability I carried as team evaluator.
Let's Talk
Need AI-ready operations support?
I design prompts, systems, and feedback loops that make AI output reliable enough for real business use.