We're hiring! This challenge is part of our recruitment process for engineering positions. We offer both remote and on-site work options to accommodate your preferences and lifestyle.
We need a chat-first, AI-powered view of how every engineer and squad are performing—both technically and in terms of business value shipped. Build a minimum-viable product (MVP) in fewer than seven days that delivers these insights inside Slack or Discord.
| Area | Requirement |
|---|---|
| Agent-centric design | Use LangChain + LangGraph agents written in Python 3.10+. Provide at least two clear roles—Data Harvester and Diff Analyst—handing off to an Insight Narrator agent via LangGraph edges. |
| Data ingestion | Pull GitHub events via REST or webhooks. The commits API exposes additions, deletions, changed_files per commit (docs.github.com); the List PR files endpoint gives the same per-file counts (docs.github.com). |
| Metrics | Track commits, PR throughput, review latency, cycle time, CI failures plus per-author diff stats (lines ±, files touched). Optionally fall back to git log --numstat for local analysis (stackoverflow.com). |
| Diff analytics layer | Your Diff Analyst agent aggregates churn, flags spikes, and links code-churn outliers to defect risk (research shows churn correlates with bugs) (stackoverflow.com). |
| AI insight layer | Agents transform data into daily, weekly, monthly narratives that map to DORA’s four keys (lead-time, deploy frequency, change-failure, MTTR) (dora.dev). Log every prompt/response for auditability. |
| Chat-first output | A Slack bot (Bolt Python SDK) (api.slack.com) or Discord bot (discord.js slash-command with embeds) (discordjs.guide) must, on /dev-report weekly, post a chart/table + the agent summary. JSON API is optional but the bot is mandatory. |
| MVP polish | One-command bootstrap (docker compose up or make run). Include a seed script with fake GitHub events so reviewers see data instantly. |
| Docs | README.md with bot install guide and an architecture diagram showing LangGraph nodes/edges, storage and chat layer. |
- Language: Python 3.10+
- Agent Frameworks: LangChain ≥ 0.1.0 (python.langchain.com) and LangGraph service or OSS package (langchain.com)
- Chat SDK: Slack Bolt-Python or discord.js (node sidecar acceptable) (api.slack.com, discordjs.guide)
- Storage: any Python-friendly store (Postgres, SQLite, DuckDB, TinyDB).
- Viz: matplotlib, Plotly, or quick-chart PNGs.
- Forecast next week’s cycle time or churn.
- Code-review “influence map” graph.
- Pluggable LLM driver (OpenAI ↔ local Llama) in < 15 min.
- Scheduled digests (bot auto-drops Monday summary).
- Pull Request to the challenge repo containing code + docs.
- ≤ 3-minute Loom/GIF demo (encouraged).
Fork today → PR in 72 hours (extensions on request). We’ll smoke-test your bot in our workspace, then book your interview.
| Category | Pts | What we look for |
|---|---|---|
| LangGraph agent architecture | 25 | Clear roles, deterministic edges, minimal hallucination. |
| MVP completeness & correctness | 25 | Metrics and diff stats accurate; bot responds; seed data works. |
| Code quality & tests | 20 | Idiomatic Python, CI green. |
| Insight value & business mapping | 15 | Narratives help leadership act. |
| Dev X & docs | 10 | Fast start, clear setup, diagrams. |
| Stretch innovation | 5 | Any wow factor. |
- Code/architecture dive (45 min)
- Edge-case & scaling Q&A (30 min)
- Product thinking & culture fit (15 min)
Original work only; public libs are fine. Don’t commit real secrets. We may open-source the winning MVP with credit.
Ready? Fork ✦ Build ✦ PR ✦ Impress us. Questions → [email protected]
- LangChain docs (python.langchain.com) – prompt, tool and memory helpers.
- LangGraph overview (langchain.com) – stateful orchestration patterns.
- GitHub Commits API (
additions/deletions) (docs.github.com) - GitHub PR Files API (per-file diff) (docs.github.com)
- Slack slash-commands guide (api.slack.com)
- Discord embeds guide (discordjs.guide)
- Git diff
--numstatusage (stackoverflow.com) - DORA four-key metrics (dora.dev)
- Code-churn research on defects (stackoverflow.com)
These resources should cover everything you need—happy hacking!