← Back to Home

Research Assessments

Marking Period 1 - Fall 2025

Assessment #1: Career Outlook

September 5, 2025

Subject: AI Research Scientist Career Path

Explored the career of an Artificial Intelligence Research Scientist, examining education requirements, job responsibilities, salary expectations, and the future of AI in various industries. Learned about the growing demand for AI skills and the diverse applications of AI technology.

Key Finding: Between 2015 and 2023, job listings requiring AI skills increased 257%, highlighting the explosive growth in the field.
Read Full Assessment

Assessment #2: Annotated Bibliography

September 12, 2025

Subject: Foundations of AI and Green Computing

Comprehensive analysis of six research sources covering green AI, AI's role in climate change, future AI predictions, privacy concerns, and medical applications. Gained foundational understanding of AI's environmental impact and the paradox of using energy-intensive technology to solve sustainability problems.

Key Insight: Some AI models require enough energy to power 121 homes for a year, highlighting the critical need for green AI development.
Read Full Assessment

Assessment #3: Deep Learning Flood Predictions

September 19, 2025

Subject: RAPFLO and Climate Resilience

Analysis of "Deep Learning Rapid Flood Risk Predictions for Climate Resilience Planning" by Yosri et al. Explored how hierarchical deep neural networks can predict floods more efficiently than traditional machine learning, using less energy while maintaining accuracy above 80% for 25-year predictions.

Technical Discovery: RAPFLO uses 24 global climate models as inputs to assess flood vulnerability and damage predictions at lower computational costs.
Read Full Assessment

Assessment #4: Air Pollution Prediction

September 26, 2025

Subject: Machine Learning in Environmental Monitoring

Examined "Air Pollution Prediction with Machine Learning: A Case Study of Indian Cities" by Kumar & Pande. Learned how machine learning models can predict air quality patterns and the challenges of data variation due to events like COVID-19. Compared energy requirements of different AI model types.

Real-World Impact: Air quality models enable cities to create green spaces and implement pollution reduction programs based on predictions.
Read Full Assessment

Assessment #5: Topic Proposal

October 2, 2025

Subject: Deep Learning in Environmental Applications

Proposed focusing ISM research on artificial intelligence and deep learning models for environmental hazard prediction. Outlined personal coding experience, mathematical background, and passion for combining AI with sustainability. Established goals for creating impactful AI models that can save lives.

Project Vision: Using deep learning to predict environmental hazards and help prepare communities for climate-related disasters.
Read Full Assessment

Marking Period 2 - Fall/Winter 2025

Research Speech Presentation

Marking Period 2, 2025

Subject: Green AI and Sustainability Research

Comprehensive presentation synthesizing research on green AI, energy consumption of large language models, and sustainable computing practices. Presented findings to class and professionals, demonstrating understanding of AI's environmental impact and potential solutions.

Presentation Focus: How prompt optimization and model efficiency can significantly reduce AI energy consumption.
View Presentation

Marking Period 3 - Winter/Spring 2026

Assessment #11: Energy-Aware Model Selection

January 16, 2026

Subject: LLM-Orchestrated AI Systems

Analysis of "Energy-Aware Data-Driven Model Selection in LLM-Orchestrated AI Systems" by Smirnova et al. Discovered how current AI orchestration frameworks waste energy through popularity-based model selection and qualitative data processing. Learned about implementing energy caps and quantitative metrics for sustainable AI.

Major Insight: Current orchestration frameworks prioritize accuracy over efficiency, choosing models that use 5% more energy for only 2% better accuracy—an unsustainable trade-off.
Read Full Assessment

Assessment #12: First Mentor Meeting

January 21, 2026

Subject: Original Work Review & Mentorship

Second interview with Ms. Marta Adamska where she agreed to become my mentor. Discussed improvements to my original work website, explored the research paper on AI orchestration systems, and received guidance on making my prompt optimizer more accessible through GitHub Pages.

Mentor Feedback: Original work successfully demonstrates prompt optimization. Should add informational pages explaining energy savings metrics to users.
Read Full Assessment

Assessment #13: LLM Energy Efficiency Trade-offs

January 30, 2026

Subject: DVFS Settings and Performance

Analysis of "Investigating Energy Efficiency and Performance Trade-Offs in LLM Inference across Tasks and DVFS Settings" by Maliakel et al. Learned about Dynamic Voltage and Frequency Scaling (DVFS) for limiting model energy consumption. Discovered that Llama model was most energy-efficient across all tasks despite being less popular than GPT.

Technical Discovery: Vague and complex prompts require significantly more energy. Energy capping can maintain acceptable accuracy while dramatically reducing consumption.
Read Full Assessment

Assessment #14: Mentor Meeting #2

February 6, 2026

Subject: Final Product Planning

Discussed final product idea with mentor: creating a green AI orchestration framework. Ms. Adamska validated the project's feasibility and impact potential. Decided to make framework open-source for developer accessibility and consider building GreenInfer chatbot as front-end demonstration.

Project Direction: Build energy-adaptive green orchestration framework with real-time energy tracking and model selection based on energy-to-accuracy ratios.
Read Full Assessment

Assessment #15: Mentor Meeting #3

February 11, 2026

Subject: Framework Architecture & Timeline

Finalized project scope with mentor. Confirmed green orchestration framework is achievable within timeline. Will prioritize framework development first, then create GreenInfer chatbot interface if time permits. Discussed implementing energy caps, real-time tracking, and automated energy profiling tools.

Next Steps: Finalize framework architecture, implement real-time energy tracking, and develop model selection algorithms for final product.
Read Full Assessment