Summary: Content strategy tailored for a PhD-seeking AI researcher (MS→Direct PhD Fall 2027, full funding). Goal: demonstrate research depth, technical competence, and communication ability to target advisors and admissions committees. Strategy: publish long-form technical deep-dives on owned domain/Substack, distribute condensed versions to LinkedIn/X, cross-link everything to wiki knowledge base.

Target Audience Priority

Priority Audience What They Need Content Format
1 PhD Advisors / Admissions Research depth, reproducibility, writing clarity Long-form blog + paper breakdowns
2 Industry Researchers Implementation insights, benchmarks, trade-offs Technical deep-dives + code
3 ML Engineers Practical tutorials, library integration How-to guides + benchmarks
4 Students / Juniors Learning paths, concept explanations Concept pages + primers
5 General Tech High-level trends, implications LinkedIn posts + threads

Content Pillars (90-Day Cycles)

Pillar 1: Efficient Training & Inference (Core expertise)

  • FlashAttention, LoRA/QLoRA, quantization
  • DSA, CSA/HCA, MLA sparse attention
  • Muon optimizer, mHC hyperconnections
  • Target venues: NeurIPS/ICML/ICLR efficiency tracks

Pillar 2: World Models & Embodied AI (Research direction)

  • Diffusion Policy, Genie, UniSim, DreamerV3
  • HRM/TRM recursive reasoning for planning
  • Sim-to-real, latent planning
  • Target advisors: DeepMind, FAIR, NVIDIA, Samsung SAIL

Pillar 3: Agentic AI & Tool Use (Engineering depth)

  • Agent SDK patterns (Claude Code, Codex Goals)
  • Context window management
  • Multi-agent orchestration
  • Demonstrates: Production-grade engineering

Pillar 4: Vision Transformers & Multimodal (Active project)

  • DSA-ViT integration (personal project)
  • HRM for visual reasoning
  • Shows: End-to-end research execution

Publishing Cadence

Platform Frequency Content Type Purpose
Owned Domain / Substack 2×/month 2,000–4,000 word deep-dives Canonical portfolio pieces
LinkedIn 2–3×/week 900–1,500 char condensed posts Professional visibility, Saves
X/Twitter 3–5×/week Threads (8–12 tweets) Real-time technical discussion
GitHub Per project Code + reproduction scripts Reproducibility proof

Rule: Every LinkedIn/X post links to canonical blog post. No orphan content.

Blog Post Types (Portfolio Pieces)

Type Length Example Purpose
Method Deep-Dive 3,000–4,000 words "Why FlashAttention Matters" Technical mastery
Paper Reproduction 2,500–3,500 words "Reproducing DeepSeek-V4" Research rigor
Tutorial + Benchmarks 2,000–3,000 words "LoRA from Scratch on RTX 4090" Engineering skill
Architecture Analysis 3,000–4,000 words "World Models Meet Diffusion" Research vision
Project Retrospective 2,000–3,000 words "Building Efficient ViT with DSA" Project ownership

PhD Application Content Checklist

Each application cycle, ensure portfolio contains:

  • 2–3 first-author paper summaries (your words, not abstracts)
  • 1–2 reproduction studies with honest limitations
  • 1 technical tutorial showing engineering competence
  • 1 research vision piece (where field should go)
  • Project pages for all personal tagged projects
  • Resume auto-generated from wiki (wiki/resume.md)

Canonical URL Hierarchy

https://ahsan.ai/blog/<slug>        ← Canonical (owned)
       ↓
https://ahsanumar.substack.com/p/<slug>  ← Newsletter import (canonical=owned)
       ↓
LinkedIn Post                       ← Condensed, link in comment
       ↓
X/Twitter Thread                    ← Section-by-section, link in reply

SEO Rule: Only owned domain gets indexed. Substack canonical → owned. LinkedIn/X noindex.

Metrics That Matter (For PhD Apps)

Metric Why It Matters Target
Substack subscribers Demonstrates audience/expertise 500+ by application
LinkedIn Saves/post Algorithm signal; reference quality 20+ per technical post
GitHub stars (project repos) Code adoption 100+ on main project
Citation/discussion Research relevance Mentions by researchers

Don't optimize for: Viral reach, generic likes, follower count.

Content Creation Workflow

Diagram: (Mermaid diagram - view source for diagram code)

graph TD
    A[Research / Paper / Project] --> B[Create/Update Wiki Concept Page]
    B --> C[Draft Blog Post in wiki/blogs/]
    C --> D[Add Mermaid, Tables, Equations]
    D --> E[Cross-link: concepts, papers, projects]
    E --> F[Export to Substack + Set Canonical]
    F --> G[Condense to LinkedIn Template]
    G --> H[Create Thread for X]
    H --> I[Publish All + Log]
    I --> J[Update wiki/resume.md]

Related Concepts

Sources