Ibrahim Ahmed

About Ibrahim Ahmed

CTO & Co-Founder @ Inference.net | Training and hosting custom AI models– 3.5 years @ Founders Inc. incubating startups in SF– Launched Leap (AI image API) → ~$1M ARR in 3 months– Built ultramail.ai (AI email assistant) → over 100k users in 6 months– Now scaling Inference.netWe see AI companies scaling with Frontier Labs facing:– Collapsing margins from 80%+ to 60-70% as Inference costs stack up– Generic models not optimized for their use case causing high latency and churn– Vendor lock-in with models that change pricing or behavior without noticeOur solutions:– Custom model training specialized for your use case– Dedicated inference infrastructure you own– Frontier-quality performance at fraction of the costEvery millisecond of GPU time = real money.I believe the future is owning specialized intelligence:Training custom models that match frontier quality for your specific use case, at economics that actually work.I post about:– Custom model training and economics– Inference optimization and GPU architecture– AI infrastructure at scale Reach out: a•••@inference.net if you\'re spending $10K+/month on LLM APIs Visit: inference.net to see how we can train and host your model

Experience

1. Co-founder, CTO

   inference.net

   May 2024 — Present · San Francisco, CA, US

   At inference.net, our team specializes in training and deploying AI models for large enterprises.We\'re powering the inference of Fortune 500 companies and fast-growing startups.Our models can completely replace leading models like GPT-5, Claude 4, and Gemini 2.5 by training models that are specialized on specific tasks. We also train models that are able to greatly enhance the performance of the frontier models using our \"Savant\" product, a method that utilizes specialized models to replace lossy embedding-based RAG pipelines with small models that are trained on vast amounts of data.Our team is able to serve these models at scale in our private cloud, on-prem, or in hybrid environments across the globe. Our proprietary inference stack achieves best-in-class performance with 99.99% uptime with blazing fast token-per-second performance.https://inference.net

Education

• Michigan State University

  Master's Degree, Computer Science - AI & Machine Learning Focus

  2018 — 2019

• Michigan State University

  Bachelor’s Degree, Computer Science

  2014 — 2018

• Edsel Ford High School

  High School

  2010 — 2014

Skills

• Microsoft Office
• Python
• Java
• C++
• C
• Git
• Kotlin
• Object Oriented Software
• Functional Programming
• Linux
• Teamwork
• Public Speaking
• Research
• Leadership
• Wordpress
• Html
• Css
• People Skills
• Persuasive Speaker
• Json
• Data Mining
• Data Science
• Machine Learning
• C#
• Javascript
• Angular2
• Typescript