Si nanostructuring by metal assisted chemical etching is highly sensitive to morphology of thin metal film, here we demonstrate use of plasmon driven optical or dielectric response of Au films for predicting nanopore or nanowire Si prior to etching, ensuring the process reproducibility. By controlling growth temperature, it is shown that same mass thickness of Au films can produce either nanopore, nanowire or hybrid (mix of the two) Si nanostructures, useful for various applications. Ag nanoparticles coated nanostructure Si is shown to be efficient surface enhanced Raman spectroscopy (SERS) substrate for dry state measurements with micro-Raman system. In contrast, by exploiting super-hydrophilicity and strong light coupling of nanostructure Si, a novel micro-drop SERS using portable Raman spectrometer is demonstrated as low-cost, reliable and easy field-deployable technique. With very low sample volume about 50 μL of as prepared Au nanoparticles and analyte, it is capable of producing SERS signal in samples containing 120 pg thiram, 240 pg Rh6G, 455 pg malachite green and 48 pg methylene blue. With advantages of micro-drop SERS, mainly high intensity signal with 3D dynamic hot-spots at high laser power, longer shelf-life colloidal gold, reusable substrates with no prior sample preparation, this method is envisaged to be an inexpensive technique for on-site trace detection applications and importantly quick generation of large experimental SERS data-set for machine learning (ML) driven studies. This is demonstrated by principal component analysis with ML classifier for 100% accurate prediction of trace dyes, pesticide and their binary mixtures.