Tanay Joshi — ECE Researcher · Distributed Systems & Networking
2nd Place Rice STEM Innovator · 1st Place ACAP Entrepreneur · Published Researcher · GPA 5.0W
The simulation crashed at 2 AM, and I had no idea why. I'd spent three weeks writing what I thought was a clean RTT-based compression controller for my drone research. The logic was sound. The math checked out. But on the actual Raspberry Pi, under real network congestion, the JPEG quality stepped down to 10% and never recovered — the feedback loop was eating itself.
That moment taught me more than any textbook. I had been thinking about the algorithm in isolation, not as something living inside a physical network with jitter, interference, and hardware limits. I had to rebuild the controller with hysteresis — a dead zone to stop it from thrashing quality levels every few hundred milliseconds. It worked. Latency dropped by up to 83%. But I never would have found that fix without first building something that broke completely.
That's the engineering philosophy I carry everywhere: build it until it fails, then understand why. I'm drawn to problems at the boundary of hardware and humanity — places where a dropped packet might mean a missed medication dose, or where a bad network handoff might be the difference between a student who gets help and one who doesn't. I don't think equity is just a policy question. I think it's an engineering problem, and I want to work on it.
Developed a dynamic JPEG compression framework that adjusts quality levels based on real-time Round-Trip Time (RTT) measurements. The system targets drone swarms relaying imagery in areas with no infrastructure — disaster zones, mountain rescues, rural medicine drops — where standard fixed-quality approaches fail under congestion.
Statistical validation using mixed-effects modeling confirmed a latency reduction ranging from 41% to 83% under degraded network conditions (p < 0.001). Hardware-tested on Raspberry Pi 5 using Linux Traffic Control (tc) to simulate real congestion scenarios.
Addressed a gap in connectivity research: most WiFi studies focus on dense urban deployments. In low-density settings (1–5 users), WiFi congestion behaves differently, and LiFi's signal confinement becomes a practical security and performance advantage.
Engineered hardware switching logic using an MCP3008 ADC and photodiode on the Raspberry Pi 5 (SPI interface via MOSI/MISO pins 19, 21, 23, 24). The system prioritizes LiFi when ambient light exceeds 1.65V. WiFi held a 20ms baseline; LiFi showed a 10% improvement in high-brightness conditions, with validated signal confinement as a physical-layer security benefit.
An AI-powered tool that helps users audit their own media consumption and flag synthetic or misleading content. Optimized for browser-side inference using TensorFlow.js so all analysis runs locally — no data leaves the device, ensuring 100% user privacy. Built on the idea that critical thinking is a skill you can engineer for.
Founded and built a budget management platform now managing financial workflows for a nonprofit theater company. The centralized MVC architecture replaced a spreadsheet-and-email system, reducing administrative overhead by around 75%. Built iteratively with the theater managers — they told me what was broken, I fixed it, repeat.
Designed, built, and presented the entire platform as sole technical architect — from database schema to the live judge demo. Connects under-resourced students with peer tutors in real time. The students most likely to need help are also least likely to have a private tutor at home; this was built for them.
Selected as 1 of 100 students for Rice University's STEM Young Scholars workshop. Led hardware design of an autonomous supply system for temperature-sensitive Rheumatoid Arthritis medication across a 200-mile Himalayan route. The core challenge: vibration at altitude degrades drug potency. Solution: a payload chamber using carbon fiber, EVA foam, and rubberized damping with triangular structural geometry to neutralize resonance frequencies.