A 3-month roadmap into ML research: from foundations to reproducing papers and contacting 15 labs
cneuralnetwork · x · 2026-10-30
A student-oriented, month-by-month plan for breaking into ML research in 3 months:
- Month 1 (foundations): brush up linear algebra, probability, calculus, optimization; get comfortable with PyTorch, autograd and training loops; implement an MLP, CNN and basic attention from scratch; reproduce a classic like ResNet on a small dataset, documenting curves and failures.
- Month 2 (reading & reproducing): pick one area (LLMs, CV, efficient training), read 5-8 papers; choose a recent paper with open code and accessible compute, reproduce its main experiment against reported numbers, and understand baselines and limitations.
- Month 3 (experiments & outreach): run 2-3 meaningful ablations, test one hypothesis-driven modification, compare across seeds, write a short technical report, open-source the code, and reach out to 10-15 relevant labs.
The author invites feedback from active researchers and PhD students.
Related event: A Roadmap for Breaking Into ML Research in 3–6 Months(2 posts)→
More from Research
- OpenAI researcher offers credits or GPU hours within 3 days for math formalization work — willdepue · 2026-10-31
- Prune-then-loop retrofit turns a pretrained VLM recurrent, beating baseline with 23.3% fewer language params — HildeKuehne · 2026-10-31
- Microsoft Research open-sources Agensh: 1,024 coding agents, no boss, 55% on Pandoc rebuild — JafarNajafov · 2026-10-31
- Tsinghua team's ActionCodec2 streams action tokens per step, hitting 46.2% grasp vs FAST's 9.2% — ZeYanjie · 2026-10-31
- OpenAI leak's hidden gem: a Log-Sobolev result completing KLS picture — IgorCarron · 2026-10-31
- "Probably present": A Visual Explainer of How Bloom Filters Work — YashasGunderia · 2026-10-31