I am a third-year undergraduate student with a background in
software engineering (Go, Java, basic frontend).
I am currently transitioning into machine learning systems,
with long-term interests in reinforcement learning infrastructure
for large language models and embodied intelligence.
I am following a self-directed 18–24 month learning path focused on
building deep foundations rather than chasing frameworks — math
first, systems thinking second, and production-quality projects as
the final output.
"Math is the dividing line between engineers who can debug and
innovate beneath the surface, and those who only know how to use
the framework."
Research Interests
- ML Systems — distributed training, GPU performance, training infrastructure
- RL Infrastructure — RLHF pipelines, post-training systems for LLMs
- Embodied Intelligence — sim-to-real transfer, robotics policies
Current Focus (Phase 1 of 4)
Building mathematical foundations and transitioning my daily
programming language from Go/Java to Python.
- Linear Algebra — reading Gilbert Strang's
Introduction to Linear Algebra; watching 3Blue1Brown's
Essence of Linear Algebra series.
- Probability & Statistics — MIT OCW 6.041
(Bertsekas), with emphasis on concepts that recur in ML
(KL divergence, MLE, cross-entropy).
- Python Fluency — NumPy, Matplotlib, and idiomatic
Python following Luciano Ramalho's Fluent Python.
Roadmap
- Phase 1 Math foundations + Python (Months 1–6)
- Phase 2 ML fundamentals + first real project (Months 7–12)
- Phase 3 Distributed training, GPU programming, RL basics (Months 13–18)
- Phase 4 Specialization + capstone portfolio project (Months 19–24)
News
- 2026-05 Started Phase 1 of the ML systems learning journey.
Projects
Hands-on Python notes and exercises following Bro Code's full-course
tutorial. Each module includes concept summaries, exercises written
from scratch (not copy-pasted), and personal debugging notes.
Python
learning
More projects coming as I progress through Phases 2–4
(planned: NumPy-only linear/logistic regression, a small Transformer
from scratch, distributed training experiments, and a capstone project
in either RLHF infrastructure or embodied RL).
Skills
- Languages I'm fluent in: Go, Java, basic frontend (React/Vue)
- Languages I'm learning: Python (NumPy, Matplotlib)
- Math: Linear algebra (in progress), probability (planned), multivariable calculus (planned)
- ML/Systems: Foundational — to be built systematically across the 24-month path
Principles I Follow
- Do, don't just watch. Implement every concept from scratch at least once.
- Math first, frameworks second. Frameworks change every 18 months; math doesn't.
- Output publicly. GitHub commits, technical writing, small contributions to open source.
- Reassess every 3 months. The plan is a draft I keep revising, not a contract.
Miscellaneous
I keep a learning journal documenting what I get wrong — the
failures are usually more instructive than the wins. I believe the
boring work (debugging, profiling, reading source code) is where the
real understanding compounds.
