Yashvardhan Singh
  • Works
  • Blog
  • About
Works

© 2026 Yashvardhan Singh · New Delhi, India

Connectmail@yashvardhan.devLinkedInCV / ResumePrivacy
  1. Home/
  2. Works/
  3. Cache Cliff edge-AI research

Systems research · 2026

Cache Cliff edge-AI research

A hardware-aware way to choose vision encoders for edge visual-language systems.

Research FellowBenchmarkingEdge inferenceLatency analysis
Visit live project ↗

Situation

Edge AI selection often prioritizes accuracy while overlooking the actual latency bottleneck.

Constraint

Recommendations needed to reflect how architectures behave on constrained hardware, not just in server benchmarks.

Decisions

Synthesized public benchmark and systems evidence into a hardware-aware selection framework for edge visual-language systems.

Outcome

Introduced the Cache Cliff terminology to make non-linear hardware trade-offs explicit; readers should validate recommendations on their target hardware.

Related Technical Research & Systems Deep-Dive

The Cache Cliff: a hardware-aware edge AI framing →

Read the empirical benchmark across 8 vision encoders and 4 hardware platforms detailing roofline limits and time-to-first-token optimization.