Faisal Azhar

Senior Technical Program Manager (Tpm) @Meta

Presently leading Generative AI Large Language Model Global programs with a goal of AI research breakthroughs. • Landed LLaMA model launch to open source 7-65B models beating GPT-3, including stepping in as PM to drive GTM strategy. • Landed a key moment in the AI & Math space with an all-in-one FLAIM conference (13 Talks across 2 days, 2 demos and 100+ ppl), VSCode plugin marketplace release (~+600% API calls/day) and panel discussion (Yann LeCun, Joelle Pineau and Sir. Timothy Grower) in Paris, France; surprising the community with our AI model that solved 12 IMO problems — 5x more than any previous AI system, beating OpenAI SOTA by 20 percent on 3x benchmarks, including NeurIPS paper. • Landed all key approvals with Legal, Privacy, Policy, Commercial counsel, Data, Code, IP and Research leads for using third party datasets (48M+ scientific corpus, 106B tokens) to train and open-source our Galactica (120B) AI model — beating Google’s Chinchilla (70B), PaLM (540B), Bigscience’s BLOOM (176B), Meta’s OPT-175B, OpenAI’s GPT-3 on Math and science tasks. • Landed 2x increased compute capacity by on-boarding 3x projects on Meta’s Research Super Cluster (RSC) with 1024, 600+ and 512 GPUs respectively for 12 months to train large language models. • Supported the Galactica AI demo website launch to publicly demonstrate the capabilities of the model by driving domain purchase and configuration setup, legal Terms of Use, obtaining 18x EC2 GPU instances under project’s AWS account for peak traffic management. Tweet (48hrs): 3M impression. • Top 10% of privacy contributors at Meta - across all privacy programs and Privacy Review Experience.