struct Sreerama {
role: "Final-year CS @ Rajarajeswari College of Engineering (VTU) · GPA 8.25",
currently: ["Java Full Stack Intern @ Ethnotech", "Building Tala App", "Veena Shruthi Analyser"],
published: "RATM — ICMCSI 2026 (IEEE)",
obsessions: ["systems programming", "Carnatic music tech", "RL for OS problems"],
fun_fact: "I write Rust. I also play veena. These are not unrelated.",
}I like problems that sit at the intersection of things that don't usually talk to each other — operating systems and machine learning, Carnatic music theory and signal processing. If it's a weird crossover, I'm probably already thinking about it.
┌─────────────────────────────────────────────────────────────────────────────┐
│ RATM: Reinforcement Learning for Co-Optimised CPU Scheduling │
│ and NUMA Memory Management in Compiler Design │
│ │
│ ▸ Presented at ICMCSI 2026 (International Conference on Mobile │
│ Computing and Sustainable Informatics) │
│ ▸ DOI: https://doi.org/10.1109/ICMCSI67283.2026.11412692 │
└─────────────────────────────────────────────────────────────────────────────┘
Built a NUMA-Aware Adaptive Tiered Allocator in Rust — three layers: Thread-Local Caches → NUMA-Node-Local Arenas → Global Page Allocator — then layered an RL agent on top to co-optimise CPU scheduling and memory placement. The kind of thing that sounds academically clean until you're debugging arena fragmentation at 2am.
🦀 Minimal Kernel Simulator with RL · Rust · Jan 2025 – Jan 2026
Simulated OS-level memory management from scratch. Designed a NUMA-Aware Adaptive Tiered Allocator that models hardware-software interaction at the memory subsystem level — thread-local caches, NUMA-node-local arenas, and a global page allocator working in concert. Applied Reinforcement Learning to co-optimise CPU scheduling and NUMA memory placement. This became the ICMCSI 2026 paper.
🎵 Tala App · Flutter + Rust · Feb 2026 – Present
A metronome that actually understands Carnatic music. Most apps count beats. This one understands Angas.
- All 72 Melakartha Talas + Sapta Talas with every Jati variation
- Mathematically precise beat timing — zero drift, because Carnatic performance demands it
- Anga-aware bouncing ball visualisation: Laghu, Dhrutha, Anudhrutha, Guru, Plutha boundaries rendered correctly
- Vilamba Kala slow-tempo mode — something no existing metronome app does
🔬 Veena Shruthi Analyser · Python · librosa · NumPy · Mar 2026 – Present
Western pitch tools are deaf to Carnatic microtones. This fixes that.
- Extracts fundamental frequency (F0) from Carnatic Veena audio via pYIN probabilistic pitch estimation
- Maps against all 22 Carnatic shrutis using just intonation frequency ratios — not the A440 equal-temperament model
- Tonic-relative model: analysis works regardless of tuning reference, which is how Carnatic music actually works
- Computes cent deviation between detected pitch and ideal shruti — exposing microtonal characteristics that no Western tool captures
🚌 RedBus Clone — Bus Booking System · Java · MariaDB · Feb 2026
Full-featured bus-booking backend built with strict layered DAO/Service architecture. Modelled entities (Bus, Route, Booking, Seat, User), wrote MariaDB DDL schemas, implemented all CRUD and booking-logic service methods. Business logic wired through JDBC. Clean separation of concerns throughout.
| Role | Company | Period |
|---|---|---|
| Java Full Stack Development Intern | Ethnotech Academic Solutions Pvt. Ltd. | Feb 2026 – Present |
| Android App Developer Intern | Prodigy Infotech | Apr 2025 |
♩ ♪ ♫ ♬ ♩ ♪ ♫ ♬
72 Melakartha Ragas. 22 Shrutis. Infinite Gamakas.
These aren't just musical concepts to me —
they're the design requirements for software nobody else thought to build.
♬ ♫ ♪ ♩ ♬ ♫ ♪ ♩
I come from a family with deep roots in Carnatic music. with my father a Veena Vidwan and scholar. Growing up surrounded by ragas, talas, and the precise mathematics of Indian classical music gave me a strange lens for computing problems. Carnatic music is extraordinarily systematic: 72 parent scales, each with defined microtonal relationships, rhythmic structures with named subdivisions. It practically asks to be modelled computationally.
The Tala App and Veena Shruthi Analyser are my attempt to give back — to build tools that actually serve this tradition rather than flattening it into Western approximations.
"The code compiles. The shruti is in tune. Both feel the same."