Let’s keep this thread idea-only — share your concept, connect with others individually if you want to team up, and once a proposal takes shape we can encourage forming a subgroup within the AI/ML AWG.
Hi everyone — we’re opening a space for concise (1–2 paragraph) project ideas that could be submitted to the Edison Scientific ACU Grant challenge.
If you have a scientific question, dataset, or hypothesis you’d like to explore, please share it here.
To keep this thread focused, please post ideas only. If you see alignment with someone else’s concept, feel free to reach out directly and collaborate organically.
What is Kosmos?
Kosmos is an AI scientist that can read literature at scale, run real data-analysis code, and produce auditable, citation-linked scientific reports. It’s designed to accelerate scientific discovery, making it a potentially powerful tool for exploring OSDR and space biology datasets. Looking forward to seeing what ideas emerge and how natural collaborations form.
For eg: Progestin Mimicking Molecule for Space Missions can deem collaboration between @FemaleReproAWG and @AIMLawg .
Best,
AI/ML AWG
@evartsb @james.casaletto @rtscott2001 @lauren.sanders @MultiOmicsAWG @ALSDAawg @HardwareAWG @BrainAWG @MicrobesAWG @AnimalAWG @PPawg @PlantAWG @HUMANawg @RLWG
6 Likes
Thanks for sharing Ms Nagesh
Can we have it as follows:
AI-Driven Carbon Control System for Data Centers Using Soft (Software) + Hard (Hardware) Absorption Mechanisms
Data centers contribute roughly 1% of global CO₂ emissions, yet real-time carbon control remains fragmented across cooling, compute, and energy sourcing. I propose a scientific study to develop and test an AI-driven Carbon Control Mechanism (CCM) that models, predicts, and actively regulates carbon emissions using both soft mechanisms (software-based AI agents that optimize workloads, cooling cycles, VM placement, and energy-mix switching) and hard mechanisms (physical micro-carbon absorbers, biochar-based filters, mineralized carbon capture plates, or algal microreactors placed within exhaust ducts). The central hypothesis is that a unified AI model can reduce net carbon output of a micro-data center pod by 18–35% through dynamic prediction and intervention.
The experiment will use sensor streams (temperature, energy draw, airflow, carbon ppm, rack load) to train a Carbon Digital Twin of the test environment, enabling the AI agent to simulate absorption efficiency, identify emission spikes in advance, and autonomously activate either software-level or hardware-level mitigation. The study will measure how well AI can optimize: (1) energy efficiency, (2) cooling load, (3) real-time carbon capture effectiveness, and (4) overall net-zero potential. Results could form the basis of a next-generation AI-for-Climate model to retrofit existing data centers with modular carbon control units.
1 Like