About the challenge
This isn't a hobby hackathon. It's how Zoftware is hiring right now for backend and AI roles, run as a 12 hour build sprint instead of a stack of interview rounds. If you've been doing this long enough to have real opinions about architecture, and you're genuinely looking for your next role, this is built for you.
Who this is for
This is strictly for experienced engineers. We're asking for 5+ years of professional backend engineering experience, with real hands on work building with LLMs, agents, or RAG pipelines, production systems or serious personal projects, not a weekend tutorial. This is not a beginner hackathon and we're not running it like one. If you're earlier in your career, hold off and wait for something built for that stage, we'd rather you not burn a Saturday/Sunday here.
The challenge: Hyper Personalised Focus 500
Build a region based software discovery feed. Think of it as a Focus 500 list, except one that can be scoped down to a single region or even a sub region, India narrowed to Delhi, for example. We'll hand you 2 to 3 launch regions to start with: UAE, India, and Global.
Requirements
What to Build
Core pieces we expect working end to end:
- Aggregation engine. Ingest software listings and community signal (reviews, forum mentions, pricing) from a seed dataset we provide per region.
- Sanitization pipeline. Dedupe entries, normalize ratings, pricing, and feature tags, and resolve conflicting data. The seed dataset is messy on purpose, this is one of the main things we're judging.
- RAG pipeline. Given a software name, pull relevant community and review context and generate a grounded summary of strengths, weaknesses, and what people are actually saying.
- Feed curator. A frontend that filters by region and shows the ranked, trending list alongside the per software analysis
If you've got time left, two things separate the top finishers from everyone else:
- An on demand scraper that pulls live data for a new software given just a URL, tested live at demo time against something we hand you.
- Geographic drill down. Can your system actually narrow a region further, India down to Delhi, say, or is the logic hardcoded to the regions you were given
What to Submit
Submissions are judged live, not through a video or write up submitted in advance. Here's what you need ready by demo time.
A working system you can run live for the judges.
A private GitHub repository. Invite [our GitHub handle will be announced live at the event] as a collaborator, do not make this repository public.
A short README covering your architecture and the tradeoffs you made along the way.
Be ready to walk us through your approach and answer questions live during the demo slot.
The actual problem statement is handed out at check in, there's nothing to prepare or build beforehand beyond registering and applying.
Prizes
Hired
Either cash or an AirPods Max 2, winner's choice, plus an offer for one of our open backend or AI engineering roles at Zoftware.
Fast-Tracked Interview
The next four strongest performers skip straight to a final round interview for open backend and AI roles at Zoftware. No resume screen, no recruiter call first.
Devpost Achievements
Submitting to this hackathon could earn you:
Judges
Lakshya
Engineering Lead @ Zoftware
Judging Criteria
-
Data sanitization, 25
how well you guys handle the messy seed data and conflicting sources. -
Retrieval and grounding, 25
whether the per-software summaries are actually supported by the context you pull. -
System design and code quality, 20
frontend polish is not scored, since this is backend first. -
Ranking and feed logic, 15
whether "trending" and "latest" mean something and change sensibly by region. -
Defense round, 15
how you handle challenge, with honesty about weak spots scoring better than bluffing. -
Bonus, upto 10
5 for the on demand scraper on an unseen URL, 5 for geographic drill down.
Questions? Email the hackathon manager
Invite others to compete
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.