B.Tech Computer Engineering β’ Vishwakarma Institute of Technology (VIT Pune)
Every major engineering system I've built maps to a celestial body in my 3D Solar System & Galaxy Portfolio (procedural WebGL simulation built with Three.js):
| Celestial Body | Flagship Engineering System | Planetary Specialty & Metric |
|---|---|---|
| βοΈ The Sun | Akash Kumar (Core) | Central Thesis: GPU Infrastructure, Bare-Metal & Low-Latency Systems |
| βΏοΈ Mercury | CipherSieve |
Sub-ms payload-agnostic encrypted TLS 1.3 threat detection (< 1.0 ms) |
| βοΈ Venus | Argus-ML |
High-pressure streaming statistical drift (KS / PSI) & DAG root-cause engine |
| π Earth & Moon | LogicVerse |
Visual IDE for Discrete Mathematics, DAG topology & Automata simulators |
| βοΈ Mars | Crimson Orbit |
Sub-15ms (11.8 ms) real-mode 8086 WebAssembly AudioWorklet synthesis |
| β Jupiter | nano-vllm |
Titan of Memory: Paged KV-Cache, continuous batching & 0.00% fragmentation |
| β Saturn | ORCA |
Distributed orchestration ring of 8 specialized marine intelligence agents |
| Metric | Measured Value | Significance | Benchmark Target |
|---|---|---|---|
| β±οΈ Audio Latency | 11.8 ms | Sub-15ms Real-Time Bare-Metal Synthesis | Crimson_Orbit |
| β‘ I/O Bus Interception | 0.889 Β΅s | Targeted Real-Mode PIT 8253 Bus Trapping | Crimson_Orbit |
| π§ VRAM Fragmentation | 0.00 % | Paged KV-Cache Dynamic Block Allocation | nano-vllm |
| π Threat Classification | < 1.0 ms | Payload-Agnostic Early-Flow TLS 1.3 Detection | CipherSieve |
| π Continuous Batching | Token-Level | Dynamic Iteration Scheduling & Zero OOM | nano-vllm |
I am a systems and AI infrastructure engineer focused on GPU memory architectures, high-throughput LLM serving runtimes, low-latency computer architecture, and bare-metal systems programming.
- π AI Systems & Deep Learning Runtimes: Engineering Paged KV-Cache memory allocators, continuous batching iteration-level schedulers, and Copy-On-Write (CoW) prefix caches for transformer inference engines.
- β‘ Low-Level Architecture & Firmware: Bare-metal 16-bit real-mode x86 assembly, custom 512-byte MBR bootloaders, bus-cycle hardware interception (PIT 8253 / PPI 61h), and WebAssembly AudioWorklet bridges.
- π‘οΈ Network Systems & Security Intelligence: Sub-millisecond payload-agnostic behavioral threat classification on encrypted (TLS 1.3 / HTTPS) traffic using 13 transport-layer temporal dynamics.
- π Formal Computation & Visualization: Interactive topological DAG engines, Turing Machine tape visualizers, pushdown automata (PDA) stack simulators, and automated counter-example finders.
- π¬ Engineering Philosophy: Designing hardware-conscious systems software, verifying microsecond telemetry, and building rich developer tools.
|
|
|
|
β‘ nano-vllmHigh-Throughput Paged KV-Cache LLM Inference Engine & GPU Memory Management Runtime
|
π΅ Crimson OrbitLow-Latency Bare-Metal 8086 Audio Synthesis & Acoustic Resynthesis Platform
|
π‘οΈ CipherSievePayload-Agnostic Sub-Millisecond TLS 1.3 Threat Classification Engine
|
ποΈ Argus-MLProduction Machine Learning Observability & Statistical Drift Engine
|
β‘ LogicVerseInteractive Discrete Mathematics & Automata Simulation IDE
|
π ORCAAgentic Marine Intelligence & Distributed Decision-Support Orchestrator
|