📱 Add me on WeChat — free pack
Open WeChat, search the ID, send 「资料」
WeChat ID
pyao2015
Send 资料 and get it free
真题精选 PDF
Study-abroad: OPT / timelines / résumé checklist
No sign-up needed: high-frequency interview points, RAG / Agent system-design essentials, and Big-Tech talent-program intel. Always-free long reads, updated continuously.
High-frequency Q&A across the full stack — Transformer internals, alignment, RAG, Agents, KV-Cache and quantization deployment — to spot your interview blind spots.
Read →Chunking, hybrid search, two-stage recall + rerank, multi-hop retrieval, hallucination & grounding, plus evaluation and monitoring approaches and engineering pitfalls.
Read →What makes an Agent different from chat, Function Calling, tools & memory, multi-agent guardrails, and how to approach Agent system-design questions.
Read →JD TGT / Alibaba Star / ByteDance Top Seed·Jindouyun / Tencent Qingyun / Huawei Genius / Meituan Beidou / Baidu AIDU — who can apply, what it takes, and the pay.
Read →US OPT timelines & visa pitfalls, autumn-recruiting eligibility for returnees, US vs. China résumé differences, and how interviews differ — everything in one guide.
Read →These guides cover the questions — but interviewers dig into your projects.
Want me to look at your résumé and tell you what to change? Message me on WeChat. Want someone alongside you all the way to the offer? See 1-on-1 job-search coaching →
Agent evaluation, LLM evaluation, reinforcement learning, RAG, multimodal — 13 real online apps. Unlimited use during your membership, full question bank and lecture notes included. Membership ¥1k/year.
An interviewer's view: what screening really looks at, how to write project experience without a laundry list, which tech-stack phrasings give you away, and how to cover a missing LLM project.
Read →When to fine-tune (vs RAG and prompting), SFT instruction data and loss curves, LoRA/QLoRA internals, RLHF vs DPO, and the memory/engineering trade-offs of full fine-tuning vs PEFT.
Read →Prefill/Decode phases, KV Cache memory math, PagedAttention and continuous batching, FlashAttention, GPTQ/AWQ/GGUF quantization, vLLM and speculative decoding — with an answer skeleton per question.
Read →An individual.
Two kinds of people — and both are included once you enrol: 1-on-1 job-search coaching is not a choice between two tracks, you get both parts together. One: your résumé does get interviews, but you keep stumbling when you have to walk an interviewer through your projects and defend the hard parts — usually because those hard parts never made it into the résumé and you cannot tell them out loud. After enrolling, I rewrite your key projects and send them back for reference, then about 1–3 months of Q&A by message to get your résumé, projects and interview prep in shape (one optional ~45-minute mock interview). Two: you are on the whole road — applying, waiting for replies, first and second rounds, all the way to the offer — and nobody is pushing the pace with you. Some do not yet have their résumé and projects ready, some get interviews but fail the technical round, some have been through several companies and no longer know what to prepare next. That part is included too: three online 1-on-1 sessions (in-depth résumé and direction alignment, an in-depth mock interview plus coaching, and a consultation), running from 3 months through to your offer, with custom learning goals and custom projects.
Not for job hunting, but for keeping pace with the AI era: staying in touch with people working at the frontier of AI, and being able to use the projects to support your own homepage — embracing AI.
1-on-1 job-search coaching ¥8,050; membership ¥1k/year
I'm firmly against freeloading off other people's time.
The interview long reads and PDF question sets stay free forever, with no sign-up; the 13 AI apps in the Project Lab need a membership — add me on WeChat. 1-on-1 job-search coaching also goes through WeChat.
📱 Scan to add me on WeChat
WeChat ID
pyao2015 Copy