Lesson 167 — Advanced

Portfolio & Positioning

Projects tab tak nahi dikhte jab tak unhe present na kiya jaaye. Yeh lesson har project ko kaise document, deploy aur explain karta hai — bina exaggerated claims ke, jo5 minute mein defend ho sakein.

Portfolio kya hai

Rule of thumb: 3 deep projects >10 todo repos. Har repo ka fate yahi hai — koi clone kare toh chalta bhi hai, README se samajh bhi aata hai, demo link kaam karta hai.

README template

# <Project name> — one-line what + for whom

![demo](screenshot-or-gif)     <- pehle screen par, text baad mein

## Live demo
<deployed URL>  (ya video link agar nahi chal raha)

## Problem
Kis pain point ke liye, real context ke saath (2-3 lines)

## Design choices (kyun aisa banaya)
- chunk size / model choice / fallback — sab trade-off ke saath

## Eval / numbers
Apne eval set se real numbers — kaunsa metric, kitne samples

## Setup
Python3.11 + `pip install -r requirements.txt` + `.env.example` copy karo

## Limits (jo nahi kiya)
Honest boundaries — yeh section interview mein trust dilata hai

• Screenshot/GIF upar: visitor ka pehla5 second yahin bit-ta hai.
• Setup ek command ka — broken README = kam score, achha code hone ke bawajood.
• Limits section optional nahi: jo nahi kiya batana credibility hai, kamzori nahi.

Case study format

Har project ke liye ek short write-up (README ya blog) — fixed order taaki interview mein bhi wahin follow karo:

• Problem: kis user ke liye, kitna bada — concrete context (manual dhoondhne mein20 min lagte the).
• Constraints: time, cost, hardware, data availability — asli engineering constraints.
• Trade-offs: fixed vs semantic chunking, small vs capable model, build vs buy — kya chuna aur kyun.
• Numbers (apne): eval set size, hit rate, latency p95, cost/day — jo bhi tune measure kiya.
• Learnings + next: kya fail hua, agla step kya — growth signal.

Demo walkthrough

• 2–3 minute structure: problem15s → live demo60s → tech/choices30s → limits aur next30s.
• Screen recording + voice (loom/jaise bhi) — README mein embed karo. Link dead ho toh YouTube unlisted + README link.
• Failure-proof demo: ek pre-warmed dataset, ek happy path, aur fallback (network gaya toh recorded video).
• Har interview se pehle demo ko dobara chala kar dekho — build itera hona chahiye, run hona chahiye.

Positioning — profile kaise set karein

Positioning ka matlab: tum kaun ho, kis kaam ke liye — sab kuch nahi. Ek line se shuru:

• LinkedIn headline: role + proof — "AI Engineer | RAG & agents in production | deployable projects" (keywords naturally, spam nahi).
• Pinned repos: best3 profile par, baaki pinned hata do — clutter impression deta hai.
• Consistency: GitHub, LinkedIn, resume — same projects, same names, same numbers.
• Community: README PRs/issues public rakho, ek short build-in-public post likho — engagement bhi signal hai.

Exercise

Common mistakes

• Fake metrics daal dena: "99% accuracy,10k users" bina evidence ke — interviewer ek sawaal mein gira dega. Sirf apne eval numbers, sample size ke saath.
• README nahi, sirf code dump: Bina README ke recruiter20 second mein chhod deta hai. Screenshot, demo link, setup — teenon mandatory.
• Har cheez ek repo mein: Course notes + projects + scripts sab mix toh portfolio ka impression kharab karta hai, aur kaam nahi dikhata. Clean folders, pinned best.

# README skeleton (copy karo, bharte jao)
## Kya banaya (1 para + screenshot)
## Live demo (URL)
## Kyun aisa design (trade-offs)
## Numbers (apne eval se)
## Kaise chalayein (setup)
## Limits (jo nahi kiya)