While increasingly powerful approaches enable investigation of transcription using small samples of RNA, approaches to investigate translational regulation in small populations of specific cell types, and/or (sub)-cellular contexts are lacking. Comprehensive investigation of mRNAs actively translated into proteins from ultra-low input material would provide important insight into molecular machinery and mechanisms underlying many cellular, developmental, and disease processes in vivo. Such investigations are limited by the large input required for current state-of-the-art Ribo-seq. Here, we present an optimized, ultra-low input “nanoRibo-seq” approach using 102 – 103-fold less input material than standard approaches, demonstrated here in subtype-specific neurons. nanoRibo-seq requires as few as 2.5K neurons, and exhibits rigorous quality control features: 1) strong enrichment for CDS versus UTRs and non-CDS; 2) narrow, distinct length distributions over CDS; 3) ribosome P-sites predominantly in-frame to annotated CDS; and 4) sufficient ribosome-protected fragment (RPF) coverage across thousands of mRNAs. As proof-of-concept, we calculate translation efficiencies from paired Ribo-seq and alkaline fragmented control libraries from “callosal projection neurons” (CPN), revealing divergence between mRNA abundance and RPF abundance for hundreds of genes. Intriguingly, we identify substantial translation of upstream ORFs in the 5’ UTRs of genes involved in axon guidance and synapse assembly. nanoRibo-seq enables previously inaccessible investigation of translational regulation by small, specific cell populations in normal or perturbed contexts.