Paper accepted to ICLR 2026: ProReGen: Progressive Residual Generation Under Attribute Correlations.
Making AI work in the real world
Serving Infernce, Surviving Production (Reliably)
Systems thinking grounded by deployment reality
Ajay is a Machine Learning Engineer and researcher focused on LLM inference, ML systems in production, and performance optimization, with over five years of experience building and scaling AI products at B2B startups.
A large part of that work involved edge-based computer vision deployments with hard hardware limits, distributed camera fleets, and the kind of operational failures that do not show up in tidy benchmark papers. That experience now informs a broader interest in inference systems, serving constraints, and model efficiency across modern ML workloads.
Current work at Rochester Institute of Technology builds on those production constraints, with emphasis on efficient inference, deployment-aware system design, and making advanced models practical in production environments.
Core sections of the portfolio
Selected sections covering publications, CV, research notes, and the broader body of work.
Research papers
Publication list generated through the existing scholar workflow.
View publicationsCV and background
Education, work history, certifications, and technical depth.
Open CVResearch notes
A curated list of papers and topics shaping the current research direction.
Browse readingsReading list
Mindset, performance, and technical books that keep feeding the work.
Visit bookshelfRecent updates
Updates on research, publications, and production-facing ML work.
First submission to NeurIPS, details will be posted soon! (Oops, rejection)
Acceptance to AWARE-AI Research Traineeship (NRT)