Abhinandan Pandey , public record

02 / 06 About , the honest version

An engineer,currently compiling.

I'm Abhinandan , a computer science undergraduate in New Delhi who builds AI products and keeps the receipts. This page is not a list of adjectives. It's how I actually operate, and why.

01 , Where this starts

B.Tech in Computer Science at AKTU, class of 2028. Within one semester it was obvious that my field and my syllabus were moving at different speeds , the curriculum treats machine learning as an elective while the industry treats it as oxygen.

I didn't drop out and I didn't complain. I run two curricula in parallel: the degree for fundamentals that don't expire, and a self-assigned one , build something real, publish it, write down what it taught me, repeat. Everything on this site is output from that second curriculum.

It started, honestly, with certificates , Microsoft and LinkedIn badges in generative AI. They taught me vocabulary and nothing else. The day I understood that difference is the day this record actually begins.

02 , The thesis

Plenty of engineers can build.Far fewer can explain.I'm training both, on purpose.

Communication is a competitive advantage in engineering , not a soft skill, a force multiplier. It's why CareAI exists (an exercise in translating medicine to humans), why I write essays, and why every case study here explains its trade-offs in plain language. The engineers I admire most are the ones whose thinking you can actually follow.

03 , How I work

01

Ship embarrassingly early

A rough thing at a URL teaches more than a polished thing in a folder. Every project here went public before it felt ready , that discomfort is the tuition.

02

Show the unflattering number

78.33% accuracy is on this site, with decimals. Calibrated honesty is rarer than competence, and it's the habit I most want to be known for.

03

Write it down, publicly

Writing is the compile step for understanding. If the explanation doesn't work, the understanding didn't either.

04

Choose rooms over lines

I pick opportunities by the people they put me next to, not the line they add to a résumé. The line is a byproduct.

05

Fundamentals don't expire

Frameworks churn every six months. Data structures, systems, statistics and clear writing keep compounding. I invest accordingly.

04 , What I'm becoming

The direction is AI engineering with product taste: LLM systems, evaluation, agents , the layer where models meet actual users. The projects are getting less like exercises and more like products; the writing is getting less like posts and more like a body of work.

I'm early, and I'd rather be honestly early than falsely senior. What I can promise anyone who works with me: momentum you can audit , it's all here, timestamped.

If you're building something where an obsessive learner who ships and writes clearly is useful , the contact page is short.

05 , Now

Building

Prompt Optimizer v2 , eval harness first

Learning

LLM evals, retrieval, agent patterns

Reading

Engineering blogs > timelines

Writing

AI education, engineering honesty

Updated July 2026 , a stale “now” section is a broken promise