A legendary island of talents.
Every company is looking in the same places. We show you the ones no one has mapped yet.
Recruiters at every company chase talent at the same ten universities. The patterns of who actually feeds whom — Plaid pulling from UPenn, Citadel from Princeton, Anthropic from Stanford — live in private group chats and LinkedIn scrolling sessions. They've never been mapped.
Talantis maps them. Every internship, every pipeline, every hidden coastline.
The product has two faces:
- The island — a clean visual surface for browsing university-to-company internship flows
- Atlas — an AI guide who walks the data with you, calls the right tools to answer your questions, and surfaces patterns you wouldn't think to ask
You can talk to Atlas anywhere. On the Talantis web app. On Agentverse. Through ASI:One. Same brain, three front doors.
"We're Stripe. Where are Plaid, Brex, and Ramp finding talent we're missing?"
Atlas calls find_similar_schools, scans 1,334 placement records across 54 companies and 31 universities, computes the gap between your peers' pipelines and yours, and returns the answer in plain prose:
"I see UPenn and NYU feeding heavily to Plaid and Brex — and you have no presence at either. That's the hidden coastline."
No SQL. No dashboards. No "let me get back to you on that." Just the answer, grounded in real data.
Three layers:
Frontend — Next.js 14, Tailwind, deployed on Vercel. The bar chart, the filters, the Atlas FAB chat panel.
Talent Intelligence Core — A Python module (api/atlas.py) that runs Claude Sonnet 4.5 in an agentic tool-use loop. Three tools defined in api/tools.py:
Recruiter-side — companies asking about pipelines
| Tool | Purpose |
|---|---|
filter_internships |
Direct factual queries — counts, lists, rankings by university × company × industry × role × year |
compare_companies |
Head-to-head pipeline comparison across multiple companies |
find_similar_schools ★ |
The signature insight — pipeline gap analysis. Where do peer companies recruit that you don't? |
Student-side — students asking about realistic options
| Tool | Purpose |
|---|---|
find_target_companies |
Companies tiered by realism — strong-fit, realistic, and reach based on your school's placement history |
analyze_school_at_company |
Your school's track record at a specific company — placement count, rank, and how peer schools fared |
discover_career_paths ★ |
The signature insight — pathway discovery. Where do students from comparable schools land that yours hasn't yet? |
Each tool maps to a Postgres function defined in schema_tools.sql for indexed performance.
Atlas uAgent — A Fetch.ai uAgent (agent/agent.py) that bridges ASI:One Chat Protocol to the Talantis backend. Same brain, different transport layer.
Atlas is registered, active, and discoverable on Fetch.ai's Agentverse marketplace.
| Property | Value |
|---|---|
| Display Name | Talantis Atlas |
| Handle | @talantis-atlas |
| Address | agent1qw9srfevfplt27z6du7xns4venmtafdezxnujl3ksamz03qwud9yurhc0hv |
| Protocol | AgentChatProtocol v0.3.0 |
| Hosted on | Render (always-on) |
| Status | Active · ASI Available |
→ View Atlas's full profile on Agentverse
Anyone with an ASI:One account can ask Atlas questions directly:
→ Live ASI:One demo conversation
Sample questions Atlas handles in stride:
- "How many UCLA students interned at Stripe last year?"
- "Compare Citadel and Jane Street's hiring patterns."
- "Which schools feed the most students to OpenAI and Anthropic?"
- "We're a fintech startup. Where should we recruit that our competitors don't?"
Multi-turn works natively — Atlas remembers what you just discussed:
You: "How many UCLA students interned at Stripe?" You: "What about Berkeley?" ← Atlas knows you mean "Berkeley → Stripe" You: "Compare those two." ← Atlas calls
compare_companiesfor UCLA + Berkeley You: "Now look at Plaid's hiring." You: "Where do they overlap with Stripe?" ← Atlas callsfind_similar_schools
Talantis's dataset covers 6,4782 internship placements across 54 companies and 31 universities for 2021-2025.
It's representative data, modeled on publicly reported placement distributions from:
- Stanford CS employment report
- Berkeley first-destination survey
- CMU SCS career outcomes
- 20+ company hiring patterns from public job postings and LinkedIn
In production, Talantis would ingest data through:
- Opt-in Chrome extension for verified placements
- University career center partnerships
- LinkedIn Talent Solutions API for enterprise customers
The hackathon dataset is realistic enough to demonstrate every Atlas capability without requiring proprietary data.
┌──────────────────────┐ ┌──────────────────────┐
│ Talantis Web App │ │ ASI:One / Agentverse │
│ (Next.js · Vercel) │ │ (Fetch.ai) │
└──────────┬───────────┘ └──────────┬───────────┘
│ POST /api/insights │ ChatMessage
│ (SSE stream) │
▼ ▼
┌─────────────────────────────────────────────────────────┐
│ Talent Intelligence Core (api/atlas.py) │
│ │
│ Claude Sonnet 4.5 ←→ 3 tools ←→ Postgres RPCs │
│ Tool-use loop · Multi-turn history · SSE streaming │
└─────────────────────────────────────────────────────────┘
│
▼
┌──────────────────────┐
│ Supabase Postgres │
│ 1,334 placements │
│ 54 × 31 grid │
└──────────────────────┘
| Layer | Tech |
|---|---|
| Frontend | Next.js 14, React, Tailwind CSS, Recharts |
| Backend | Python 3.12, FastAPI, Anthropic SDK 0.49 |
| Database | Supabase (Postgres 15) with custom RPCs |
| Agent | Fetch.ai uAgents 0.24, Chat Protocol v0.3.0 |
| Hosting | Vercel (frontend + API), Render (uAgent) |
Talantis ("tuh-LAN-tis") = legendary island of talents. The name does double duty: a play on Atlantis, and a portmanteau of talent + atlantis.
The two faces:
- Talantis — the island, the world, the visual product
- Atlas — the titan who maps the island, the AI guide
Voice: Mythic, not mystical. Curious, not certain. Spare, not sparse. Warm, not casual.
Palette: Navy #0a1628 · Gold #d4a548 · Cream #f5ecd7 · Aqua #4fb3bf (60/30/8/2)
Type: Cormorant Garamond italic for display (the gold "a" in the wordmark is the brand signature), Inter Tight for body.
talantis/
├── app/ Next.js 14 App Router
│ ├── layout.js
│ ├── globals.css
│ ├── page.jsx Landing
│ ├── explore/page.jsx Main product page
│ └── submit/page.jsx
│
├── components/
│ ├── Nav.jsx
│ ├── Footer.jsx
│ ├── TalantisLogo.jsx
│ ├── AtlasLogo.jsx
│ ├── chart/ Recharts bar chart + filter
│ └── atlas/ AtlasFAB + AtlasPanel
│
├── api/ Python serverless functions (Vercel)
│ ├── index.py FastAPI entrypoint
│ ├── atlas.py Tool-use loop + SSE streaming
│ ├── tools.py 3 tools, Postgres-backed
│ └── database.py Supabase client
│
├── agent/ Fetch.ai uAgent (deployed on Render)
│ ├── agent.py Chat Protocol implementation
│ ├── requirements.txt
│ ├── render.yaml
│ └── README.md Published to Agentverse
│
├── public/ Static assets, favicons, hero images
└── ... Tailwind, Next, Vercel config
git clone https://github.com/Talantis/talantis
cd talantis
# Frontend
npm install
npm run dev # http://localhost:3000
# Backend (in a separate terminal)
cd api
pip install -r requirements.txt
# set SUPABASE_URL, SUPABASE_KEY, ANTHROPIC_API_KEY
uvicorn index:app --reload --port 8000
# uAgent (in another terminal)
cd agent
pip install -r requirements.txt
export AGENT_SEED_PHRASE=$(openssl rand -hex 32)
python agent.py
Built for LA Hacks 2026 — Fetch.ai Agentverse: Search & Discovery of Agents track.
Deliverables:
- ASI:One shared chat session → asi1.ai/shared-chat/YOUR_UUID_HERE
- Agentverse profile → agentverse.ai/agents/details/agent1qw9srfevfplt27z6du7xns4venmtafdezxnujl3ksamz03qwud9yurhc0hv/profile
- Public repo → github.com/Talantis/talantis
- Demo video → Devpost submission
The map exists. It just hasn't been drawn yet.