DataFyno

Inspiration

Every data analyst knows the pain. A notification pops up on Slack: "Hey, can you pull the sales numbers for the last campaign?" It looks simple, but it's a nightmare. Which campaign? Which region? Gross or net sales? By "last," do they mean Q3 or the Halloween promo? We realized that 50% of a data analyst's job isn't writing SQL-it's acting as a translator. We waste hours in the "Clarification Loop," emailing back and forth just to understand what the stakeholder actually wants. We built Datafyno to end that loop forever.

What it does

Datafyno is the first AI Ambiguity Detector for data teams. It acts as a firewall between vague human requests and precise database queries. Unlike standard chatbots that hallucinate an answer based on a guess, Datafyno is programmed to stop when it detects missing context.

1. It Analyzes: It scans the request for missing dimensions (Time, Metric, Filter).

2. It Detects Ambiguity: It calculates an Ambiguity Score (A) based on the sum of weighted missing variables:

   A = Σ (Weight × Missing_Variable)

   Where "Weight" is the importance of the missing metric (e.g., a missing 'Date Range' has a higher weight than a missing 'Region').

3. It Pushes Back: If \(A > 0\), it automatically drafts a polite, professional clarification email.

4. It Solves: Once clarified, it generates the perfect, clean SQL specification.

   How we built it

   We wanted the experience to feel "fast" and "premium,".

5. Frontend: We used Next.js for a snappy, reactive UI and TailwindCSS for a sleek, dark-mode-first aesthetic.

6. Animations: To convey the "AI Magic," we heavily utilized Framer Motion. The UI pulses, glows, and transitions smoothly to visualize the AI "thinking" and "detecting."

7. AI Engine: The core logic is powered by Google Gemini. We engineered a specialized system prompt that instructs the model to prioritize identifying gaps over providing answers. It is tuned to be skeptical and precise.

Challenges we ran into

The biggest challenge was tuning the AI to not be helpful. Most LLMs try to be "nice" and guess what you mean. We had to fight that training. We had to embrace "Prompt Engineering for Skepticism"-teaching the model that guessing is a failure and that asking a clarifying question is a success. Balancing the UI was also tricky. We wanted it to look futuristic but still be functional for a serious data workflow.

Accomplishments that we're proud of

We're proud of the three-column output: Ambiguities (red), Questions (yellow), and the auto-generated Push-Back Email (purple). It visually tells the story of "Detect → Clarify → Act" in a single glance.

What we learned

We learned that sometimes the best AI feature is knowing when to stop. Most products try to do too much. We focused on one thing: catching what's missing.

What's next for Datafyno

1. Directly integrate with Slack bots to intercept vague requests before they even reach the analyst.
2. Connect to real database schemas (Snowflake/BigQuery) so the "Data Spec" creates valid, runnable SQL instantly.

Built With

• framer-motion
• google-gemini-api
• next.js
• tailwindcss
• typescript