I study applications of machine learning and artificial intelligence for physical layer waveform and receiver design in wireless and optical communications.
Designing waveform bases that reorganize how symbols interact in fractional delay–Doppler channels, so a receiver can focus on a limited number of dominant interactions.
Comparing learned designs with OTFS and AFDM under matched channels and receiver budgets to understand where adaptation helps.
Jointly learning transmit and receive bases to reduce interference from residual CFO, alongside a model-guided receiver that refines symbol estimates.
The work explores the tradeoff between communication reliability and PAPR.
Learning constellation geometry and receiver decisions together under LED clipping constraints, including modulation orders beyond the usual powers of two.
This allows finer control of transmission rate in VLC systems.
Using cross-stitch neural networks to recover multiple information streams within a composite VLC waveform, sharing useful features between related demodulation tasks.
An interpretable measure of task dependence helps explain when that shared processing is useful.
IEEE MILCOM 2026Code
IEEE ICMLCN 2026
Applied OpticsVol. 643, Issue 29, pp. 8716–8721Code
Communications & Networks ConnectVolume 1
IEEE ICMLCN 2026
University at Albany — SUNY Albany, NY, USA
University at Albany — SUNY Albany, NY, USA
Punjab Engineering College Chandigarh, India
University at Albany — SUNY Albany, NY, USA
In collaboration with IBM, developing learned waveforms for fractional delay–Doppler channels and robustness to residual CFO, with receiver-aware design and evaluation against conventional baselines.
IEEE Transactions on Communications
University at Albany — SUNY Albany, NY, USA
University at Albany — SUNY Albany, NY, USA
Returned to support preparation for Honors Orientation 2025 and related Honors College activities.
Kolligate Seattle, WA, USA Remote
Busywizzy Gurugram, India
Quantela Bengaluru, India