Cost-Effective Spin Coater
A benchtop spin coater with closed-loop speed and acceleration control, built from catalogue parts for about ₹3,000.
Arduino · Embedded C · PID control · Gain scheduling
From Last LightIndia
Siddhesh SaliAbout
Hardware6 drawn before they were built
Software5 projects, mostly ML
Photographs61 frames across 5 albums
Engineer · Photographer
Siddhesh Sali
I build instruments, write code, and make photographs.
Reach for a word to hold its frame.
Or just ask.
The whole site is a few keystrokes away. ls shows what is here,open projects takes you there, and search looks through all of it. help lists the rest — most of it, anyway.
This is a prompt when JavaScript is on. Without it, everything it would have told you is on the rest of this page: the work is below, the photographs after that, and there is a search page.
Selected work
0311From the circuit
to the code.
Instruments, embedded systems, and machine learning. Selected projects from the workbench and the lab.
Language Guidance for Deep Metric Learning
Using pretrained language models as a supervisory signal for vision embeddings — a multi-stage schedule that fine-tunes intermediate layers rather than only the head.
PyTorch · Deep metric learning · Vision-language models · ResNet-50
Power Amplifier for B–H Characterisation
A transformer-less, direct-coupled linear amplifier that drives magnetic cores from 50 Hz to 250 kHz, so their B–H loops can be measured across four decades of frequency.
Analogue design · KiCad · Proteus · MATLAB
Photographs
5 albumsA reason to
look closer.
High mountain roads, familiar fields, and the life in between. Landscape, wildlife, and travel photographs since 2015.
Rural India
Work, water and weather — eight frames of people getting on with the day, made over four years in Maharashtra.
Field Notes
Birds and small things, almost all at 300mm — mostly campus and the scrub around it, over three years.
Ladakh
Three days between Leh, Nubra and Pangong — high desert, a lake that changes colour by the hour, and the marmots that own the grass.