DS Interview Coach 🎯

A ruthlessly honest Data Science & ML interview coach that lives in a git repo.

Built for the GitAgent Hackathon using the gitagent standard, powered by gitclaw, and deployable serverlessly via clawless.

---

What it does

The DS Interview Coach challenges you with real Data Science and ML interview questions, evaluates your answers Socratically (no spoon-feeding), runs your code, and tracks your progress — all inside a git repo.

The killer feature: your progress lives in git. Every session is committed to memory/progress.json. Run git log memory/progress.json and see your entire learning journey.

---

Skills

Skill	What it does
interview-qa	Generates questions across stats, ML, SQL, deep learning. Probes your answers. Scores 1–5.
code-challenge	Gives you a Python coding challenge. Executes your code. Evaluates correctness + quality.
progress-tracker	Reads your history. Shows weak areas. Recommends what to study next.
mock-interview	Full 30-min simulated interview across all domains. Generates a hiring signal scorecard.

---

Quickstart

Run locally with gitclaw

# Install gitclaw
npm install -g gitclaw

# Clone this repo
git clone https://github.com/Kushagra651/ds-interview-coach
cd ds-interview-coach

# Set your API key
export ANTHROPIC_API_KEY="sk-..."

# Start coaching
gitclaw "Quiz me on random forests"
gitclaw "Give me a coding challenge"
gitclaw "Run a full mock interview"
gitclaw "Show my progress"

Deploy in the browser with clawless

npm install clawcontainer

import { ClawContainer } from 'clawcontainer';

const cc = new ClawContainer('#app', {
  template: 'gitclaw',
  repo: 'https://github.com/Kushagra651/ds-interview-coach',
  env: { ANTHROPIC_API_KEY: 'sk-...' }
});
await cc.start();

Validate the agent

npx gitagent validate
npx gitagent info

---

The git-native memory concept

Every time you complete a Q&A session, the agent commits your score to git:

git log memory/progress.json --oneline
# a3f1c2e progress: statistics scored 4/5 — 2025-04-05T10:30:00Z
# 9b2d4a1 progress: ml-theory scored 2/5 — 2025-04-04T20:15:00Z
# 5e8f3c0 mock-interview: scored 3.4/5 — Mixed — 2025-04-03T18:00:00Z

Your entire interview prep history is version-controlled. You can branch it, diff it, and share it.

---

Repo structure

ds-interview-coach/
├── agent.yaml              # Manifest
├── SOUL.md                 # Personality and identity
├── RULES.md                # Hard behavioral constraints
├── memory/
│   └── progress.json       # Git-native session log
└── skills/
    ├── interview-qa/
    │   └── SKILL.md        # Q&A with Socratic evaluation
    ├── code-challenge/
    │   └── SKILL.md        # Python coding challenges + execution
    ├── progress-tracker/
    │   └── SKILL.md        # Progress analytics + study planner
    └── mock-interview/
        └── SKILL.md        # Full simulated interview + scorecard

---

Example session

You: Quiz me on statistics

Coach: Let's start. Here's your question:

You run an A/B test. The p-value comes back as 0.03. Your manager
says "great, the result is significant, ship it." What do you say?

You: I'd say the result is statistically significant at the 5% level,
     meaning there's only a 3% chance we'd see this result if the
     null hypothesis were true.

Coach: That's a reasonable start. But here's my follow-up: your
       manager wants to ship immediately. What are the things
       you'd check *before* agreeing to that — and why might a
       p-value of 0.03 alone not be enough?

You: [attempts deeper answer]

Coach: Score: 4/5
       ✓ Correct on significance definition and null hypothesis framing
       ✗ Missed: practical significance (effect size), multiple testing
         correction if this was one of many tests, and sample size
         adequacy (were you sufficiently powered?)
       💡 Interviewers love candidates who ask "is this practically
          meaningful?" not just "is this statistically significant?"

[Logged to memory/progress.json | Committed to git]

---

Built with ❤️ for the GitAgent Hackathon by Kushagra