Vishy Poosala — Co-Founder
I left Meta's GenAI organization to build infrastructure that makes AI deployments safer and more dependable. Inside and outside Meta, I saw the same problems delay AI projects repeatedly: - AI quality degrades silently—teams spend valuable time hunting down what changed - Compliance becomes difficult — shipping delayed because someone used the wrong data for the wrong purpose Key missing piece? Lineage infrastructure that tracks what actually happened. I'm building SEVzero—opensource metadata backbone that connects your data, models, code, and evals into one traceable graph. When your AI drifts, you see exactly what changed, you fix it in hours instead of weeks. When auditors ask questions, you have answers. And you enforce compliance policies before they ask! I also work with select companies on Dependable AI Architectures—helping teams instrument their systems, build audit trails, and create the technical foundation for trustworthy AI. Not policy frameworks—actual code, infrastructure, and implementation. BACKGROUND - Senior Engineering Director & Distinguished Engineer, GenAI @ Meta (2021-2024) - Built LLaMA platform infrastructure and safety systems - Director, Messenger Platform @ Meta - 1.8B+ users - Head of Bell Labs India - 50+ patents in AI/ML systems - 30+ research publications WHAT I BUILD - Lineage infrastructure (metadata graphs connecting AI systems end-to-end) - Drift detection systems (automated root cause analysis) - Audit trail infrastructure (who accessed what data, when, and why) - Quality monitoring (evals that run continuously, not just at release) - Safety instrumentation (making AI systems observable and debuggable) WRITING I publish weekly on "Derisking AI" (Substack)—practical insights on building safer AI systems, debugging production issues, and lessons from Meta-scale infrastructure. WORKING WITH ME I work with companies building AI in production—helping instrument systems, design audit infrastructure, and solve the gnarly debugging problems that come with scale. Typical engagement: 2 weeks where we instrument your system, build the lineage graph, set up drift detection, and create a technical roadmap you can actually execute. Best fit: Teams deploying AI in production who need to debug faster, comply with regulations, or just understand what their system is doing. If you're dealing with AI drift, compliance requirements, or mysterious quality regressions, let's talk. --- 25+ years building products and teams | Still write code daily | Based in Saratoga, CA
Stackforce AI infers this person is a SaaS-focused AI infrastructure expert with extensive leadership experience.
Location: Saratoga, California, United States
Experience: 29 yrs 11 mos
Skills
- Ai Safety
- Ai Infrastructure
- Engineering Leadership
- Product Development
- Platform Development
Career Highlights
- 50+ patents in AI/ML systems.
- Led engineering for Meta's LLaMA platform.
- Building open-source infrastructure for trustworthy AI.
Work Experience
SEVzero.ai
Founder & Happy Coder (1 yr 6 mos)
Meta
Distinguised Engineer & Senior Director of Engineering - GenAI (1 yr 6 mos)
Senior Director of Engineering - Health & Well-being (2 yrs 10 mos)
Director, Engineering - Facebook Messenger (promoted to Senior Director) (5 yrs 9 mos)
Director Of Engineering - Internet.org (8 mos)
Alcatel-Lucent
CTO - Subscriber Data Management & Messaging (11 mos)
Head of Bell Labs - India (5 yrs)
Founder & CTO - Geopepper.com, Alcatel Lucent Venture on LBS (2 yrs 10 mos)
HeadStart Ventures LLP
Mentor Partner (3 yrs)
Net Synergetics
Founder and First Hacker (2 yrs 11 mos)
Bell Labs, Alcatel-Lucent
Technical Manager (5 yrs 10 mos)
Member of Technical Staff (6 yrs)
New York University
Visiting Professor (6 mos)
IBM
Summer Intern - DB2 Query Optimizer (7 mos)
Education
Doctor of Philosophy (Ph.D.) at University of Wisconsin-Madison
Master of Science (MS) at University of Wisconsin-Madison
B.Tech at Indian Institute of Technology, Madras