| <!DOCTYPE html> |
| <html lang="en"> |
| <head> |
| <meta charset="UTF-8" /> |
| <meta name="viewport" content="width=device-width, initial-scale=1.0"/> |
| <title>Arif Adito — AI Researcher & Business Growth Leader</title> |
| <style> |
| body { |
| font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', sans-serif; |
| max-width: 800px; |
| margin: 0 auto; |
| padding: 40px 24px; |
| color: #1a1a2e; |
| background: #f8f9ff; |
| line-height: 1.7; |
| } |
| h1 { font-size: 2rem; margin-bottom: 4px; } |
| h2 { font-size: 1.3rem; border-bottom: 2px solid #e0e0f0; padding-bottom: 8px; margin-top: 40px; } |
| h3 { font-size: 1.1rem; color: #3a3a7a; } |
| .subtitle { color: #555; font-size: 1rem; margin-bottom: 4px; } |
| .location { color: #888; font-size: 0.95rem; margin-bottom: 32px; } |
| .badge { |
| display: inline-block; |
| background: #1a1a2e; |
| color: white; |
| padding: 6px 14px; |
| border-radius: 20px; |
| font-size: 0.85rem; |
| text-decoration: none; |
| margin: 4px; |
| } |
| .badge:hover { background: #3a3a7a; } |
| ul { padding-left: 20px; } |
| li { margin-bottom: 6px; } |
| .paper-box { |
| background: white; |
| border-left: 4px solid #3a3a7a; |
| padding: 16px 20px; |
| border-radius: 8px; |
| margin: 16px 0; |
| box-shadow: 0 2px 8px rgba(0,0,0,0.06); |
| } |
| .model-card { |
| background: white; |
| border: 1px solid #e0e0f0; |
| padding: 16px 20px; |
| border-radius: 8px; |
| margin: 12px 0; |
| box-shadow: 0 2px 8px rgba(0,0,0,0.04); |
| } |
| .model-card h4 { margin: 0 0 8px 0; color: #1a1a2e; } |
| .model-card .specs { color: #666; font-size: 0.9rem; } |
| .stack { color: #555; font-size: 0.95rem; } |
| .certs { display: flex; flex-wrap: wrap; gap: 8px; margin-top: 8px; } |
| .cert { |
| background: #eef0ff; |
| color: #3a3a7a; |
| padding: 4px 12px; |
| border-radius: 12px; |
| font-size: 0.8rem; |
| } |
| .links { margin-top: 40px; } |
| .footer { margin-top: 60px; text-align: center; color: #aaa; font-size: 0.9rem; } |
| .highlight { color: #3a3a7a; font-weight: 600; } |
| .role { color: #555; font-size: 0.9rem; margin-top: 2px; } |
| </style> |
| </head> |
| <body> |
|
|
| <h1>Arif Adito - @Adiuk</h1> |
| <p class="subtitle">AI Researcher | Founder, Adioris Tech Ltd | Head of Business @ Tapmad Bangladesh</p> |
| <p class="location">Dhaka, Bangladesh</p> |
|
|
| <p>Strategic leader with 15+ years scaling ventures across SaaS, OTT, and Fintech. Now building sovereign AI infrastructure for underserved markets — custom models, no dependencies, consumer hardware.</p> |
|
|
| <h2>Research & Models</h2> |
|
|
| <h3>Noor: Custom Sparse MoE Language Model</h3> |
| <p><em>No Python. No PyTorch. No Adam. Built from scratch in Rust.</em></p> |
|
|
| <div class="model-card"> |
| <h4>Noor Model Family</h4> |
| <div class="specs"> |
| <strong>Noor-Edge</strong> (2.8B) — Phone, RPi, browser. PLE architecture, 64K Borno tokenizer.<br/> |
| <strong>Noor-Pro</strong> (12B, 3B active) — Laptop. Parallel Dense+MoE, Sigmoid+SMEBU routing, Sandwich RMSNorm.<br/> |
| <strong>Noor-Max</strong> (28B, 4B active) — Workstation. Full architecture with Block AttnRes. |
| </div> |
| </div> |
|
|
| <ul> |
| <li><span class="highlight">MuonClip optimizer</span> — Muon + QK-Clip. 67% less memory than Adam.</li> |
| <li><span class="highlight">SMEBU routing</span> — Sigmoid routing with momentum bias balancing.</li> |
| <li><span class="highlight">Parallel Dense+MoE</span> — Dense FFN + expert specialization per token.</li> |
| <li><span class="highlight">Borno tokenizer</span> — Custom 64K BPE for English + Bangla + code.</li> |
| <li><span class="highlight">First bilingual English+Bangla model</span> at this scale built without PyTorch.</li> |
| </ul> |
|
|
| <a class="badge" href="https://huggingface.co/Adiuk/noor-edge-checkpoints" target="_blank">Checkpoints</a> |
| <a class="badge" href="https://huggingface.co/datasets/Adiuk/noor-training-data" target="_blank">Training Data</a> |
|
|
| <h3>Eyla: Identity-Anchored LLM Architecture</h3> |
|
|
| <div class="paper-box"> |
| <strong>arXiv:2604.00009</strong><br/> |
| "Eyla: Toward an Identity-Anchored LLM Architecture with Integrated Biological Priors — Vision, Implementation Attempt, and Lessons from AI-Assisted Development"<br/><br/> |
| <a class="badge" href="https://arxiv.org/abs/2604.00009" target="_blank">Read on arXiv</a> |
| <a class="badge" href="https://github.com/Adiuk24/eyla-aios" target="_blank">GitHub</a> |
| </div> |
|
|
| <ul> |
| <li>Identity-anchored fine-tuning (Soul Pass)</li> |
| <li>HiPPO-initialized SSMs for biologically-inspired memory</li> |
| <li>Identity Consistency Score (ICS) — novel benchmark</li> |
| </ul> |
|
|
| <h2>Projects</h2> |
|
|
| <div class="model-card"> |
| <h4>Eyla AIOS</h4> |
| <p class="specs">Local-first agentic AI OS. Noor as the core model, ADE verification pipeline (P1-P6), multi-device deployment. Targeting 130M+ users via Grameenphone and Robi Axiata partnerships.</p> |
| </div> |
|
|
| <div class="model-card"> |
| <h4>AdiTurbo</h4> |
| <p class="specs">Ternary quantization (TQ3) for edge deployment. ARM NEON kernels. 4.5x compression with minimal quality loss.</p> |
| <a class="badge" href="https://huggingface.co/Adiuk/AdiTurbo-TQ-Models" target="_blank">Models</a> |
| </div> |
|
|
| <div class="model-card"> |
| <h4>Adioris Tech Ltd</h4> |
| <p class="specs">AI solutions, digital transformation, and strategic consulting. Building sovereign AI infrastructure for Bangladesh and emerging markets.</p> |
| </div> |
|
|
| <h2>Experience</h2> |
| <p><strong>Head of Business, Bangladesh</strong> — Tapmad</p> |
| <p class="role">Led market entry strategy (0 to 1). GTM, telecom partnerships (Grameenphone, Robi Axiata), revenue growth.</p> |
|
|
| <p><strong>Founder & CEO</strong> — Adioris Tech Ltd</p> |
| <p class="role">AI research, custom model development, strategic consulting.</p> |
|
|
| <p><strong>CEO & Director of Photography</strong> — Studio By Adi (London)</p> |
| <p class="role">Commercial photography and cinematic production.</p> |
|
|
| <h2>Certifications</h2> |
| <div class="certs"> |
| <span class="cert">Strategic Leadership — Harvard Business School</span> |
| <span class="cert">Advanced Growth Strategy — Reforge</span> |
| <span class="cert">Digital Marketing Analytics — MIT Sloan</span> |
| <span class="cert">AI for Business Strategy — Wharton</span> |
| <span class="cert">Google Data Analytics Professional</span> |
| </div> |
|
|
| <h2>Tech Stack</h2> |
| <p class="stack">Rust | Burn | Zig (NEON kernels) | GGUF | Borno 64K BPE | Metal + CUDA | RunPod | Kaggle</p> |
|
|
| <h2>Connect</h2> |
| <div class="links"> |
| <a class="badge" href="https://arifadito.com" target="_blank">Website</a> |
| <a class="badge" href="https://github.com/Adiuk24" target="_blank">GitHub</a> |
| <a class="badge" href="https://linkedin.com/in/arif-adito-025088b4" target="_blank">LinkedIn</a> |
| <a class="badge" href="https://arxiv.org/abs/2604.00009" target="_blank">arXiv Paper</a> |
| </div> |
|
|
| <p class="footer">Built in Dhaka, Bangladesh — for the world.</p> |
|
|
| </body> |
| </html> |
