Project Title
OMNI Project — Affordable AI Compute Platform for Students #IdeaInnovationBusiness
OMNI Project — Affordable AI Compute Platform for Students #IdeaInnovationBusiness
Category
Computer Science
Computer Science
Authors
alainlarouche188@gmail.com
alainlarouche188@gmail.com
Short Description
OMNI: affordable student-owned AI compute infrastructure with learn-by-operating pedagogy. Phase 1 execution plan for the TeraOpenScience Science Challenge. #IdeaInnovationBusiness
OMNI: affordable student-owned AI compute infrastructure with learn-by-operating pedagogy. Phase 1 execution plan for the TeraOpenScience Science Challenge. #IdeaInnovationBusiness
Long Description
# OMNI Project — Phase 1 Execution Plan ## TeraOpenScience "Idea, Innovation, Business: A Science Challenge" — #IdeaInnovationBusiness **Team Lead:** Alain Larouche (Québec City, Canada) **LinkedIn:** https://www.linkedin.com/in/alain-larouche-7613b41b6/ **Team:** Solo entry (open to adding 1–2 members before October 31, 2026) **Date:** October 2026 --- ## 1. The Innovation **OMNI is an affordable AI compute platform built for students.** Students who want to do serious AI work — training small models, running inference experiments, learning how LLMs actually run — hit the same wall: the cost of compute. Consumer GPUs are unaffordable on a student budget, and cloud GPU offerings are priced for enterprises, not for a student running their first fine-tuning job at 2 AM. OMNI's innovation is a **student-owned, pedagogical AI infrastructure model**: real physical compute nodes, orchestrated as one accessible platform, paired with an educational layer that teaches while it computes. Instead of renting opaque cloud instances, students get hands-on access to real hardware, real queues, real trade-offs — the way engineers actually learn infrastructure. The novel application is twofold: 1. **Democratized AI compute**: turning affordable/consumer-grade hardware into a shared, student-accessible AI platform — proving that meaningful AI experimentation doesn't require enterprise budgets. 2. **Learn-by-operating pedagogy**: the platform itself is the classroom. Users don't just run models; they see how scheduling, VRAM management, and distributed inference work on real machines. ## 2. Executive Summary OMNI Project builds affordable AI infrastructure for students. Phase 1 (this entry) covers the execution roadmap. The project already has working physical infrastructure: four server nodes in Québec City, including one NVIDIA RTX 3060 12GB dedicated to AI workloads. Over the five-month Phase 2 build period (Nov 2026 – Apr 2027), the team will: (1) harden a job orchestration layer across the four nodes, (2) ship a simple student-facing web interface for submitting AI jobs, (3) run a pilot with a first cohort of student users, (4) document a proof-of-concept with measured results, and (5) deliver a full business plan. The project requires no revenue to date to be meaningful: it is infrastructure-first, education-driven, and already operational at small scale. Grant funds would accelerate hardware capacity and cover operating costs during the build. ## 3. Team **Alain Larouche — Team Lead (solo)** - Finishing a multidisciplinary bachelor's degree, Université Laval (December 2026) - Founder of OMNI Project; building and operating the 4-node infrastructure solo - Team lead in healthcare operations (weekends) — experience running small teams under pressure - Daily technical content producer (LinkedIn, Instagram, X, Threads), bilingual podcast producer, YouTube channel operator — demonstrates ability to communicate complex technical topics simply - LinkedIn: https://www.linkedin.com/in/alain-larouche-7613b41b6/ *Note: solo entries are scored on the same criteria; the team is open to adding 1–2 members (e.g., a backend developer or a UX contributor from the Université Laval network) before the October 31 roster lock to strengthen delivery capacity.*
# OMNI Project — Phase 1 Execution Plan ## TeraOpenScience "Idea, Innovation, Business: A Science Challenge" — #IdeaInnovationBusiness **Team Lead:** Alain Larouche (Québec City, Canada) **LinkedIn:** https://www.linkedin.com/in/alain-larouche-7613b41b6/ **Team:** Solo entry (open to adding 1–2 members before October 31, 2026) **Date:** October 2026 --- ## 1. The Innovation **OMNI is an affordable AI compute platform built for students.** Students who want to do serious AI work — training small models, running inference experiments, learning how LLMs actually run — hit the same wall: the cost of compute. Consumer GPUs are unaffordable on a student budget, and cloud GPU offerings are priced for enterprises, not for a student running their first fine-tuning job at 2 AM. OMNI's innovation is a **student-owned, pedagogical AI infrastructure model**: real physical compute nodes, orchestrated as one accessible platform, paired with an educational layer that teaches while it computes. Instead of renting opaque cloud instances, students get hands-on access to real hardware, real queues, real trade-offs — the way engineers actually learn infrastructure. The novel application is twofold: 1. **Democratized AI compute**: turning affordable/consumer-grade hardware into a shared, student-accessible AI platform — proving that meaningful AI experimentation doesn't require enterprise budgets. 2. **Learn-by-operating pedagogy**: the platform itself is the classroom. Users don't just run models; they see how scheduling, VRAM management, and distributed inference work on real machines. ## 2. Executive Summary OMNI Project builds affordable AI infrastructure for students. Phase 1 (this entry) covers the execution roadmap. The project already has working physical infrastructure: four server nodes in Québec City, including one NVIDIA RTX 3060 12GB dedicated to AI workloads. Over the five-month Phase 2 build period (Nov 2026 – Apr 2027), the team will: (1) harden a job orchestration layer across the four nodes, (2) ship a simple student-facing web interface for submitting AI jobs, (3) run a pilot with a first cohort of student users, (4) document a proof-of-concept with measured results, and (5) deliver a full business plan. The project requires no revenue to date to be meaningful: it is infrastructure-first, education-driven, and already operational at small scale. Grant funds would accelerate hardware capacity and cover operating costs during the build. ## 3. Team **Alain Larouche — Team Lead (solo)** - Finishing a multidisciplinary bachelor's degree, Université Laval (December 2026) - Founder of OMNI Project; building and operating the 4-node infrastructure solo - Team lead in healthcare operations (weekends) — experience running small teams under pressure - Daily technical content producer (LinkedIn, Instagram, X, Threads), bilingual podcast producer, YouTube channel operator — demonstrates ability to communicate complex technical topics simply - LinkedIn: https://www.linkedin.com/in/alain-larouche-7613b41b6/ *Note: solo entries are scored on the same criteria; the team is open to adding 1–2 members (e.g., a backend developer or a UX contributor from the Université Laval network) before the October 31 roster lock to strengthen delivery capacity.*
Potential Applications
## 4. Project Plan
### 4.1 Team roles
- **Alain Larouche (Team Lead):** architecture, backend/orchestration, infrastructure operations, user onboarding, reporting.
- **(If recruited) Member 2:** web interface development.
- **(If recruited) Member 3:** documentation and user testing coordination.
### 4.2 Milestones — Phase 2 build (Nov 16, 2026 → Apr 23, 2027)
| Period | Milestone | Deliverable |
|---|---|---|
| Nov 16 – Dec 15, 2026 | Orchestration layer v1 | Job queue running across 4 nodes; GPU scheduling on the RTX 3060; basic monitoring and logging |
| Dec 16, 2026 – Jan 15, 2027 | Student interface v1 | Simple web UI: submit inference/fine-tuning jobs, view queue status, read guided tutorials |
| Jan 16 – Feb 15, 2027 | Pilot cohort | First 10–20 student users onboarded (recruited from Université Laval peer network); usage feedback collected |
| Feb 16 – Mar 15, 2027 | Iteration + proof-of-concept | Measured results documented (jobs run, success rate, user feedback); cost-per-job analysis |
| Mar 16 – Apr 23, 2027 | Business plan + defense | Full business plan with financial projections; final defense preparation |
Progress reports to the jury in the weeks of January 18 and March 15, 2027, per the challenge schedule.
### 4.3 Technical development strategy
- **Honest baseline:** the infrastructure exists today — four physical nodes, one with an RTX 3060 12GB for AI. No invented capacity.
- **Stack (open-source):** Linux on all nodes; containerized workloads; an open-source inference server for LLM workloads; a lightweight job queue (e.g., Redis-backed) with a small custom scheduler prioritizing fair student access; a minimal web frontend.
- **Development approach:** incremental, node by node. B1 (GPU node) first for inference jobs; B2 as entry/API gateway; B3 for orchestration logic; B4 (NAS) for model and dataset storage.
- **Testing:** each milestone ends with a documented test (e.g., "10 concurrent inference jobs complete with median latency" — targets to be set from measured baselines, not invented).
- **Risk mitigation:** single-GPU bottleneck is acknowledged — the scheduler design and the pilot's small cohort size are explicitly sized to it. Hardware failure contingency: jobs are stateless and re-queueable.
### 4.4 Supply chain roadmap
- **Already owned:** all four server nodes, the RTX 3060 12GB, networking gear. Zero supply-chain dependency for Phase 2's core build.
- **If granted ($5,000–$7,500):** a second GPU (same class, ~12GB VRAM) to double AI capacity; additional RAM/storage for the NAS; a managed switch upgrade; electricity and hosting costs during the build period. All commodity, off-the-shelf parts — no exotic sourcing.
- **Scaling beyond Phase 2:** the model scales by adding refurbished enterprise nodes (a well-established secondary market), keeping per-unit costs student-compatible. No custom silicon, no proprietary dependencies.
## 4. Project Plan
### 4.1 Team roles
- **Alain Larouche (Team Lead):** architecture, backend/orchestration, infrastructure operations, user onboarding, reporting.
- **(If recruited) Member 2:** web interface development.
- **(If recruited) Member 3:** documentation and user testing coordination.
### 4.2 Milestones — Phase 2 build (Nov 16, 2026 → Apr 23, 2027)
| Period | Milestone | Deliverable |
|---|---|---|
| Nov 16 – Dec 15, 2026 | Orchestration layer v1 | Job queue running across 4 nodes; GPU scheduling on the RTX 3060; basic monitoring and logging |
| Dec 16, 2026 – Jan 15, 2027 | Student interface v1 | Simple web UI: submit inference/fine-tuning jobs, view queue status, read guided tutorials |
| Jan 16 – Feb 15, 2027 | Pilot cohort | First 10–20 student users onboarded (recruited from Université Laval peer network); usage feedback collected |
| Feb 16 – Mar 15, 2027 | Iteration + proof-of-concept | Measured results documented (jobs run, success rate, user feedback); cost-per-job analysis |
| Mar 16 – Apr 23, 2027 | Business plan + defense | Full business plan with financial projections; final defense preparation |
Progress reports to the jury in the weeks of January 18 and March 15, 2027, per the challenge schedule.
### 4.3 Technical development strategy
- **Honest baseline:** the infrastructure exists today — four physical nodes, one with an RTX 3060 12GB for AI. No invented capacity.
- **Stack (open-source):** Linux on all nodes; containerized workloads; an open-source inference server for LLM workloads; a lightweight job queue (e.g., Redis-backed) with a small custom scheduler prioritizing fair student access; a minimal web frontend.
- **Development approach:** incremental, node by node. B1 (GPU node) first for inference jobs; B2 as entry/API gateway; B3 for orchestration logic; B4 (NAS) for model and dataset storage.
- **Testing:** each milestone ends with a documented test (e.g., "10 concurrent inference jobs complete with median latency" — targets to be set from measured baselines, not invented).
- **Risk mitigation:** single-GPU bottleneck is acknowledged — the scheduler design and the pilot's small cohort size are explicitly sized to it. Hardware failure contingency: jobs are stateless and re-queueable.
### 4.4 Supply chain roadmap
- **Already owned:** all four server nodes, the RTX 3060 12GB, networking gear. Zero supply-chain dependency for Phase 2's core build.
- **If granted ($5,000–$7,500):** a second GPU (same class, ~12GB VRAM) to double AI capacity; additional RAM/storage for the NAS; a managed switch upgrade; electricity and hosting costs during the build period. All commodity, off-the-shelf parts — no exotic sourcing.
- **Scaling beyond Phase 2:** the model scales by adding refurbished enterprise nodes (a well-established secondary market), keeping per-unit costs student-compatible. No custom silicon, no proprietary dependencies.
Open Questions
### 4.5 Business plan outline (for Phase 2)
1. **Problem & customers:** students and student-adjacent builders priced out of AI compute.
2. **Solution:** affordable, education-oriented AI compute access on real infrastructure.
3. **Revenue model (to be validated in Phase 2):** freemium tier for students (free quota, paid bursts); institutional licenses for student clubs, labs, and schools. No revenue figures are projected at this stage — pricing will be set from the pilot's measured cost-per-job.
4. **Go-to-market:** the existing content channels (6 platforms, podcast, YouTube) are the acquisition engine; the pilot cohort becomes the first testimonial base.
5. **Financials:** 10-year projections and NPV to be developed during Phase 2 from real operating data — not invented for this entry.
## 5. Real-world impact
- **Scientific:** gives students hands-on access to AI infrastructure, producing practitioners who understand compute — a bottleneck skill in AI research.
- **Societal:** lowers the economic barrier to AI education; a student in Québec City gets the same experimentation power as one with a corporate cloud budget.
- **Commercial:** validates a low-cost infrastructure model that could serve schools, bootcamps, and student organizations — a market cloud providers price out.
## 6. University policy compliance
*Statement (to be verified by the Team Lead before posting):* OMNI Project is an independent personal initiative. It does not use Université Laval computing resources, funding, or facilities; development happens on privately owned hardware. The Team Lead will confirm compliance with Université Laval's policies on student entrepreneurship and intellectual property before the October 31 deadline, and will note any required disclosures in the posted entry.
## 7. Grant use ($5,000–$7,500, if awarded)
- Second GPU (~12GB VRAM class): doubling AI job throughput
- NAS storage/RAM expansion: model and dataset hosting
- Network upgrade (managed switch): inter-node reliability
- Operating costs (electricity, domain, hosting) for the 5-month build
- All spending documented in the January and March progress reports, per the challenge's installment schedule (50% kickoff / 25% / 25%).
---
*Prepared October 2026. To be posted in the PROJECT section of the TeraOpenScience platform with the tag #IdeaInnovationBusiness before October 31, 2026.*
### 4.5 Business plan outline (for Phase 2)
1. **Problem & customers:** students and student-adjacent builders priced out of AI compute.
2. **Solution:** affordable, education-oriented AI compute access on real infrastructure.
3. **Revenue model (to be validated in Phase 2):** freemium tier for students (free quota, paid bursts); institutional licenses for student clubs, labs, and schools. No revenue figures are projected at this stage — pricing will be set from the pilot's measured cost-per-job.
4. **Go-to-market:** the existing content channels (6 platforms, podcast, YouTube) are the acquisition engine; the pilot cohort becomes the first testimonial base.
5. **Financials:** 10-year projections and NPV to be developed during Phase 2 from real operating data — not invented for this entry.
## 5. Real-world impact
- **Scientific:** gives students hands-on access to AI infrastructure, producing practitioners who understand compute — a bottleneck skill in AI research.
- **Societal:** lowers the economic barrier to AI education; a student in Québec City gets the same experimentation power as one with a corporate cloud budget.
- **Commercial:** validates a low-cost infrastructure model that could serve schools, bootcamps, and student organizations — a market cloud providers price out.
## 6. University policy compliance
*Statement (to be verified by the Team Lead before posting):* OMNI Project is an independent personal initiative. It does not use Université Laval computing resources, funding, or facilities; development happens on privately owned hardware. The Team Lead will confirm compliance with Université Laval's policies on student entrepreneurship and intellectual property before the October 31 deadline, and will note any required disclosures in the posted entry.
## 7. Grant use ($5,000–$7,500, if awarded)
- Second GPU (~12GB VRAM class): doubling AI job throughput
- NAS storage/RAM expansion: model and dataset hosting
- Network upgrade (managed switch): inter-node reliability
- Operating costs (electricity, domain, hosting) for the 5-month build
- All spending documented in the January and March progress reports, per the challenge's installment schedule (50% kickoff / 25% / 25%).
---
*Prepared October 2026. To be posted in the PROJECT section of the TeraOpenScience platform with the tag #IdeaInnovationBusiness before October 31, 2026.*
AI Assistant
Please review this Phase 1 execution plan against the challenge jury criteria (innovation and originality 25%, technical feasibility 25%, execution plan quality 25%, team composition 15%, real-world impact 10%) and suggest concrete improvements to strengthen the entry before the October 31 deadline.
Please review this Phase 1 execution plan against the challenge jury criteria (innovation and originality 25%, technical feasibility 25%, execution plan quality 25%, team composition 15%, real-world impact 10%) and suggest concrete improvements to strengthen the entry before the October 31 deadline.
Keywords
IdeaInnovationBusiness
IdeaInnovationBusiness
Email
alainlarouche188@gmail.com
alainlarouche188@gmail.com