Inspiration
High-value opportunities are everywhere—hackathons, jobs, consulting engagements, grants, scholarships, accelerators, partnerships, local opportunities, and more. The problem is no longer finding links. The harder problem is deciding which opportunities are actually worth pursuing.
Users still have to ask: Is this current? Am I eligible? Does it fit my goals and capabilities? Is the potential upside worth the effort? Have I already evaluated it? Has anything changed? And what should I do next?
Opportunity Hunter was created to close that gap between discovery and action.
What it does
Opportunity Hunter turns opportunity discovery into a five-stage workflow:
Find → Qualify → Prioritize → Track → Act
A user describes what they want to find in ordinary language. The agent searches current sources, verifies candidate-specific evidence, evaluates opportunities against the user's goals and constraints, ranks them using a structured scoring framework, remembers prior evaluations, and generates pursuit plans for the strongest opportunities.
Instead of returning another list of search results, Opportunity Hunter creates a prioritized opportunity portfolio with evidence, decisions, and next actions.
How I built it
Opportunity Hunter is built as an agentic workflow using Google Agent Development Kit (ADK) and Gemini 3.6 Flash on Vertex AI.
Google Search provides current opportunity discovery. The system then verifies sources, normalizes and deduplicates candidates, and applies deterministic Python scoring across five dimensions: strategic fit, capability fit, expected value, probability of success, and effort efficiency.
Cloud Firestore provides persistent memory. This allows repeated Hunts to identify opportunities as NEW, CHANGED, or UNCHANGED, rather than forcing the user to evaluate the same information repeatedly.
A specialist Pursuit Strategist generates an immediate next action and practical pursuit plan for strong opportunities.
The application is served through FastAPI and publicly deployed on Google Cloud Run.
Challenges
One of the biggest challenges was source quality. Live search can easily find generic organization pages, directories, or stale information. I added source guardrails so strong recommendations require evidence tied to the specific opportunity whenever such a source should exist.
Another challenge was repeated work. Finding the same opportunity again is not useful unless the system understands whether something changed. Persistent Firestore state and change detection were added to solve that problem.
I also found that AI-generated pursuit strategies could become temporally incorrect—for example, recommending planning stages that had already passed. The strategist now receives the current date so its recommendations remain time-aware.
Finally, I had to connect the browser experience cleanly to the ADK agent workflow, which led to a server-side FastAPI execution layer.
Accomplishments
The result is a working, publicly deployed end-to-end product that supports both Saved Hunts and free-text Hunts, personalized scoring, live source discovery, persistent memory, change detection, PURSUE / CONSIDER / PASS recommendations, detailed analysis, pursuit plans, and CSV export.
The system has been tested against strong-fit opportunities, poor-fit opportunities, eligibility failures, duplicate candidates, missing information, exact-source verification, and repeated Hunts that exercise NEW → UNCHANGED → CHANGED behavior.
What I learned
The biggest product insight was that discovery is becoming commoditized. AI can find things very quickly.
The more valuable layer is decision infrastructure: understanding the user, enforcing evidence quality, separating signal from repetition, making tradeoffs explicit, and turning the strongest opportunities into concrete actions.
Technically, I also learned that combining probabilistic AI agents with deterministic scoring, structured schemas, source rules, and persistent memory produces a more trustworthy workflow than relying on a single open-ended model response.
What's next
The next stage is to turn Opportunity Hunter from an on-demand agent into an ongoing opportunity-management platform: recurring Hunts, notifications only when meaningful opportunities appear or change, résumé/profile ingestion, Gmail/Calendar/Sheets integrations, collaborative Hunts, outcome tracking, and learning from the opportunities users actually pursue.
Opportunity Hunter's product principle: Don't give me more things to search. Give me the opportunities worth acting on.
Development note
The current implementation was developed beginning August 24, 2026, during the AI Builders Hackathon build period. The same implementation was also submitted to the Google Cloud All Things Agentic Hackathon. This submission focuses on Opportunity Hunter's broader product, workflow, technical implementation, usability, and real-world potential.## Inspiration
Log in or sign up for Devpost to join the conversation.