aviv sheriff
Motif·AI Portfolio·Aviv Sheriff
Project 01 of 06·Jun — Nov 2025

Breakout

Breakout maps thousands of real game launches to extract the patterns behind hits — and turns them into tailored pitch validation and go-to-market plans.

Product Gaming B2B Marketing Tech Social Listening Founder INACTIVE
01Artifacts
02At a glance
Role
FounderSolo
Timeline
5 monthsJun — Nov 2025
Codebase
~35k LOC3 repos · 410 commits
Stack
Next.js · FastAPIPostgreSQL · Multi-LLM
03The work

Game developers — from indie solo teams to mid-size studios — decide what to build next with gut feel. Breakout turned that decision into a data product, built as three complementary modules.

Pitch evaluation and greenlighting — an 8-step evaluator that grades a new game concept against the patterns of thousands of real launches (differentiation, market timing, art style fit, audience match, execution plausibility), on top of a searchable library of launch stories. GTM planning — a timeline library of how successful games actually went to market (community seeding, content creator reach, demo timing, launch cadence), turned into tailored campaign plans. Micro-tools — a Steam page scoring engine, a tag-network momentum analyzer, a 64-style art taxonomy, and an LLM-driven pitch-to-tags classifier, each solvable as a standalone product.

Under the hood: a 40GB+ warehouse of game data scraped continuously from Steam, SteamDB, Twitch, TikTok, YouTube, Reddit, and Twitter, filtered from 264k raw games down to ~15k high-signal titles, served by six independently-deployed Railway services. Go-to-market was high-touch: 60+ validation calls across studios and publishers including SVP-level conversations at Ubisoft, surfacing a $1M valuation offer before walking away on stage-fit grounds. Post-mortem published on LinkedIn.

04Skills
Market intelligence Competitive analysis Social listening Product strategy Customer discovery60+ calls Enterprise salesSVP-level Multi-modal LLM evaluation Semantic search Predictive modeling Data pipeline engineering40GB+ Microservices architecture Full-stack product design
05Notable accomplishments
01
Pitch evaluator and greenlighting system
The core product surface: an 8-step evaluator that takes a new game concept and grades it against a searchable library of real launches — differentiation, market timing, art-style fit, audience match, execution plausibility. Designed so a dev can go from raw idea to GTM-ready pitch in a single sitting. Built on semantic retrieval across an indexed launch-story database (ChromaDB + Cohere rerank + OpenAI Responses API), with an LLM pitch-to-tags classifier that constrains matches to Steam’s tag vocabulary. scripts/chromadb_implementation/pitch_analyzer.py
02
Steam page scoring engine
A standalone tool that grades any live Steam store page on eight rubrics — capsule, trailer, tags, screenshots, about section, short description, reviews, related games — proposes specific improvements, and normalizes the score against peer games in the same genre. Implemented as a multi-modal LLM router that assigns each component to the right model: GPT-o3 for copy-heavy pieces, Gemini 2.5 Pro with direct video upload for hero trailers, GPT-4.1 for visual tags and screenshots. steam_page_analyst/scoring/components/
03
40GB game-market data warehouse
A continuously-updated dataset of ~15k high-signal games (filtered from 264k raw scrapes), built from six public sources — Steam, SteamDB, Twitch, TikTok, YouTube, Reddit — feeding into analysis layers: a tag-network momentum engine (Jaccard + PMI + link prediction, temporally validated), a 64-style art taxonomy applied across 11k+ games, and a revenue model validated against actual Steam top-seller rankings. The data moat under every other module. GameMarketer/services/