AI research and deployment company — creator of GPT and ChatGPT
1. Inferred ICP Pre-generated
COMPANY SIZE: 5–500 employees, Seed–Series C, $500K–$100M ARR
IDEAL ROLE: CTO, Head of AI, ML Engineer, or Founding Engineer at AI-native SaaS, developer tools, or knowledge work automation companies
INDUSTRY: Developer tools, B2B SaaS, legal tech, HR tech, edtech, content platforms — any vertical integrating LLMs into core workflow
TRIGGER SIGNALS: Shipping first LLM-powered feature, switching from Anthropic/Cohere due to cost or capability mismatch, raising a round with AI-native product as core thesis
2. Sample Prospects 5 profiles
Theo Harrington
Head of AI
Clauseflow · 11–50
"Building contract review AI for mid-market legal teams — 40% accuracy improvement in 6 months"
🎯 Tweeted a GPT-4 benchmark comparison — evaluating models for a document reasoning task
Priyanka Mehta
Founding Engineer
Voicepath · 1–10
"AI voice coaching product — ships weekly, obsessively benchmarks model latency"
🎯 GitHub shows migration from Whisper self-hosted to API — cost optimization at scale
Lucas Fernandez
CTO
Synapse Labs · 51–200
"Adding AI copilot features to an existing SaaS product used by 2,000 companies"
🎯 Series B press release explicitly calls out GPT-4 integration as flagship new feature
Hana Kimura
ML Engineer
Respondly · 11–50
"Training fine-tuned support models on top of GPT base — hitting latency walls"
🎯 Posted on HackerNews about GPT fine-tuning costs — evaluating alternatives
Anton Reyes
VP of Engineering
Draftly · 51–200
"AI writing assistant for enterprise marketing teams — 50K daily active users"
🎯 Job posting for "LLM Infrastructure Engineer" — scaling their inference pipeline for enterprise reliability
Sample illustrative profiles — run a live teardown of your company to get your own.
3. Outreach Angle Ready
POSITIONING ANGLE
The pitch isn't "use GPT-4" — every engineer already knows it exists. The angle is trust and reliability at production scale. AI founders are burned by model degradation, silent API changes, and rate limit surprises in prod. Lead with stability, SLAs, and the platform ecosystem (fine-tuning, embeddings, Assistants API) as a unified substrate vs. stitching together providers.

SUBJECT LINE
Your LLM stack decision is bigger than the model benchmark

OPENING (first 2 sentences)
Noticed you're shipping your first LLM feature — the benchmark you're optimizing for in dev is almost never the bottleneck that bites you in production. Teams at your stage usually hit rate limits, context window edge cases, or silent model behavior shifts within the first 30 days of real traffic.
4. Signals to Monitor 5 signals
1. GITHUB: New dependencies on openai, anthropic-sdk, or langchain added to a previously non-AI repo → LLM integration just started
2. JOB BOARDS: "AI Engineer" or "LLM Engineer" first hire → company is going AI-native for the first time
3. TWITTER/X: Founder tweets about model comparison, GPT-4 vs Claude results, or "we switched to X" → active vendor evaluation
4. LINKEDIN: "We just shipped AI-powered [feature]" announcement → product is live, infrastructure decision locked or re-evaluatable
5. CRUNCHBASE: AI-native company raises Seed or Series A → new capital, first time building for production scale

More teardowns