WHAT I'M DOING NOW

Last updated: January 2026

THIS_WEEK.log

$ tail -f progress.log

[2026-01-26 20:00]
Implementing FlashAttention from scratch.
6 hours debugging positional encoding.
Finally works. ✓

[2026-01-25 14:30]
Reading "Attention Is All You Need" for
the 5th time. Finally understanding why
layer normalization placement matters.

[2026-01-24 09:00]
Started Building LLM Reasoners course.
Greg Durrett is incredible.

█

CURRENT FOCUS

COURSE.md

Building LLM Reasoners

Spring 2026 · Professor Greg Durrett · NYU

Learning transformer architecture from first principles
Implementing RLHF and chain-of-thought reasoning
Understanding how frontier models actually think
Not just API calls—the real mechanics

PROJECTS.json

Active Projects

{
  "h1b_rag": {
    "status": "ongoing",
    "desc": "Deterministic eval for legal Q&A"
  },
  "pico_interp": {
    "status": "ongoing",
    "desc": "10M param model interpretability"
  },
  "transformers": {
    "status": "ongoing",
    "desc": "Building from scratch, no libs"
  }
}

READING.txt

Currently Reading

"Attention Is All You Need" - Vaswani et al.
"Language Models are Few-Shot Learners" - GPT-3 paper
"Constitutional AI" - Anthropic

INTERNSHIP SEARCH

TARGET_ROLES.txt

→ Forward Deployment Engineer (Palantir-style)
→ AI Product Engineer (0→1 products)
→ Technical roles at AI startups

REQUIREMENTS.txt

What I'm Looking For

⚡ Build products that ship to real users
🚜 Work in messy 0→1 problem spaces
🤖 Move fast

STATUS: ACTIVELY LOOKING FOR SUMMER 2026

GOALS THIS MONTH

GOALS.md

[ ] Complete FlashAttention implementation
[ ] Finish H-1B RAG evaluation framework
[ ] Submit 10 internship applications
[ ] Write technical blog post on transformer internals

This page updates every Sunday.

← BACK HOME