Stirixi | Devpost

Landing page
Landing page (2)
Integrations
Team View (List of employees)
Detailed view (individual employee)
Statistics and Soulbound token information
Project View (List of projects)
Detailed Project View (including suggested hires)
Prospective Hire View
MongoDB Database
Snowflake Data (Raw data, before processing)

Inspiration

CTOs and engineering managers face a constant blind spot: who on their team is truly net positive? GitHub commits, PR reviews, and AI tool usage live in silos, making it impossible to measure real engineering efficiency. A single misaligned hire can derail a startup’s momentum, yet there’s no transparent, data-driven way to quantify performance across human and AI-assisted work.

We built Stirixi AI ATL to unify that signal. It’s the first full-stack observability platform that blends GitHub + Linear delivery data, Codex/Claude/Cursor prompt usage, Snowflake ML-driven analytics, and Solana-based proof-of-work credentials into one “CTO operating system.”

What it does

Stirixi AI ATL shows engineering leaders, at a glance, which team members drive impact and which create drag.

Live Engineering Analytics - A unified Next.js dashboard tracking PR throughput, bug rates, review latency, and AI spend across your entire toolchain
StirixiAI Copilot - A persistent Gemini-powered assistant that reads your org snapshot and recommends staffing or mentorship moves in plain language
On-Chain Reputation - Performance snapshots mint as non-transferable Soulbound Tokens on Solana Devnet, creating tamper-proof proof-of-work credentials
Full Data Stack - MongoDB (OLTP), Snowflake (OLAP), and FastAPI services feed the UI, with built-in data connectors that work out of the box

How we built it

Frontend: Next.js 16 App Router with TypeScript, Tailwind 4, Recharts, and an embedded Gemini 2.5 Flash copilot widget
Backend: FastAPI + async MongoDB exposing CRUD endpoints
Analytics Layer: ETL jobs normalize data from GitHub, Linear, and AI tools into Snowflake, where ML notebooks compute reliability, AI efficiency, and bug propensity metrics
Blockchain Layer: Solana Token-2022 program anchors performance hashes as non-transferable SBTs
Deployment: Dockerized stack with GitHub Actions and docker-compose for one-click deployment

Challenges we ran into

A big challenge for our project was setting up the ETL pipeline that pumped data into our Snowflake database. We’re pulling from a variety of sources (Github, Cursor, Linear, etc) that all require their own specific setup and attention to detail, which made things take longer than expected. Even once we had this data together, we had to make a lot of decisions about how we were going to use the intelligent systems that Snowflake provides to consolidate our data into useful metrics. Not only did we need to know what metrics were most important for determining an engineer’s performance, but we also needed to know how we were going to use the raw, messy data we had from our sources and the Snowflake intelligence to produce coherent results.

Accomplishments that we're proud of

Delivered an end-to-end analytics dashboard from scratch, complete with an AI copilot and blockchain credentials
Implemented ETL pipeline to import data from several sources into a structured Snowflake database
Minted live Soulbound Tokens on Solana Devnet with working non-transferable enforcement
Achieved a fully reproducible demo where every engineer profile, wallet, and SBT signature works without network flakiness
Shipped production-ready architecture using enterprise tools (Snowflake, MongoDB, FastAPI, Next.js 16)
Embedded a Gemini-powered copilot that translates raw metrics into formulated suggestions

What we learned

Snowflake connects analytics and blockchain proofs in ways traditional warehouses can't - it serves as both ML feature store and on-chain data source. Solana's Token-2022 extensions unlock reputation primitives that work histories can't capture; credentials should be earned, not traded. Engineering KPIs must evolve for human-AI hybrid teams, requiring new statistical models that account for both contributions. Executive AI assistance needs strict guardrails - Gemini can translate metrics to strategy only with rigorous factual grounding. The future of employee efficiency isn't more dashboards; it's unified intelligence.

What's next for Stirixi

Predictive Hiring: Train models on performance fingerprints to identify high-potential candidates pre-hire (currently using Gemini API)
Snowflake Pipelines: Incorporate data imports from a wider variety of sources to enable more employers to effectively utilize our app
Live Integrations: Replace demo connectors with production GitHub, Linear, and AI tool APIs