Baman: AI brief analyzer for freelance orders
Clients on Azerbaijani freelance platforms often post vague orders such as "I need a website, make it cheap, 20 AZN". Freelancers then spend days in chat clarifying scope, price and deadline, or accept an unrealistic job and end up in a dispute. Some briefs are also scams (off-platform payment requests, requests for ID or card data, "free sample first"). Baman is for two people: the client who cannot write a good brief, and the freelancer who has to read it. A client types the order in free text (Azerbaijani, Russian or mixed). The system returns a structured scope, milestones with payment shares, a realistic AZN price range with a confidence level, the questions still missing from the brief, and risk flags. The client gets a clearer order in under a minute instead of several rounds of messages.
Preliminary score
- Value for the user17.0 / 25
A specific problem and a clear outcome from solving it.
- Prototype and use of AI15.0 / 30
A working core scenario, and what AI actually contributes to the result.
- Quality testing4.0 / 20
Tests, examples of failures, and comparison with the current approach.
- Feasibility11.0 / 15
Data requirements, running costs and a clear next step.
- Originality5.5 / 10
A meaningful difference from common solutions.
What the judges wrote
- GPT-6 AstraPreliminary50/100
- Strengths
- Your focus on unclear Azerbaijani freelance briefs identifies a concrete pain for both clients and freelancers. Your combination of scope, payment milestones, missing questions and scam signals is well matched to that workflow. Your materials also address server-side secrets, quota limits and hosting cold starts, showing practical awareness of operating constraints.
- To improve
- Your submission needs a captured /analiz input-to-output run; the current materials describe the scenario without demonstrating its result, while the README foregrounds a broader Lumi marketplace assistant. Replace the testing placeholders with actual counts, pass rates, latency, token usage and concrete failed outputs. Measure clarification time or message count against freelancers' current workflow rather than assuming one analysis resolves the missing information. Validate AZN estimates against completed orders or expert quotes, and make their uncertainty explicit.
- Claude Opus 5.5Preliminary55/100
- Strengths
- Your problem framing is concrete and local: vague orders like "a website for 20 AZN", disputes over unagreed scope, and scam patterns such as off-platform payment and ID requests are real pains for Azerbaijani freelancers. The AI is essential to the result, turning messy Azerbaijani, Russian or mixed text into structured scope, milestones, an AZN price range with confidence, missing questions and risk flags. Your guardrails are sensible: a strict JSON schema, unknown details routed to a missing-info list instead of being invented, and server-side keys with rate limiting. Your feasibility slide is clear on data, cost and next step, and your disclosure of the reused Freelancer.az UI shell is transparent.
- To improve
- Your quality-testing section was submitted as an unfilled template: [N], [X]/[N], [Y] seconds and [Z] tokens are placeholders, and the listed failures are only hypothetical examples. As a result, there is no evidence of actual pass rates or real failure cases. Fill in real numbers from your results file, list two or three concrete failed briefs with the model's actual output, and time a real before/after clarification to support the 'one request instead of several messages' claim. Showing even one full example output in the deck would also make the working prototype verifiable, and your README should be aligned with this submission, since it describes a broader 'Ailancer/Lumi' marketplace assistant.
The team
İlahə Nəsirli
business
Adil Dostəli
developer
Vidadi NəsirliLead
developer