Vasanth Balakrishnan
Member of Technical Staff (Founding Research Engineer) at Resolve AI
About
I'm Vasanth Balakrishnan, a Founding Research Engineer at Resolve AI. My career has been dedicated to the intersection of machine learning and robust engineering, from building real-time identity verification systems as a founding engineer at Vouched to developing ML infrastructure at Humane. Currently, I'm focused on creating 'AI Production Engineers'—agentic systems capable of autonomously troubleshooting and resolving complex production issues. I am passionate about moving engineers from being 'in the loop' to 'on the loop,' allowing AI to handle mission-critical, long-horizon tasks. With a background in Engineering Physics from IIT Madras and deep experience in LLMs and computer vision, I offer expertise in building scalable, purpose-built AI models. I'm always looking to connect with high-caliber talent and industry partners who are interested in the future of automated operations and AI-driven observability.
Networking
What I can offer
- ›Deep expertise in building agentic systems and ML infrastructure
- ›Insights into identity verification and fraud detection research
- ›Experience scaling AI startups from founding stages
Looking for
- ›expanding my professional network
- ›exploring mutual opportunities in AI and Production Engineering
Best fit for
Current Interests
Background
Career
Began as a software developer intern at Mad Street Den, transitioned to freelance ML consulting, then served as Head of AI Research at Vouched. Later worked as a Senior ML Infra Engineer at Humane before becoming a founding research engineer at Resolve AI.
Education
Bachelor of Technology (B.Tech.) in Engineering Physics, Indian Institute of Technology (IIT), Madras (2010 – 2014)
Achievements
- ›Founding Engineer at Vouched and Resolve AI
- ›Core team member at Resolve AI during $1.5 billion valuation milestone
- ›Led research for identity verification used by major banks and healthcare institutions
- ›Developed player and ball tracking systems for US Open professional tennis teams
Opinions
- General-purpose frontier models fall short for production engineering; they require domain-specific post-training.
- As software writing becomes automated, the primary challenge shifts to operating and debugging it.
- Engineers should move from being 'in the loop' to 'on the loop,' with AI handling long-horizon tasks.