PRINCIPAL SYSTEMS ARCHITECT / DELHI, INDIA

Hi, I'm Sourav.
I architect high-performance systems.

Bridging low-level system architecture and ergonomics. I engineer browser-native computing runtimes, distributed LLM routing pipelines, and zero-latency interfaces built for scale.

HIGH-PERFORMANCE RUNTIMESZERO ARCHITECTURAL BURN
Build for
throughput.
From low-level data structures
to seamless human interaction.
S / 2026CODE
→
SCALE
Scroll to explore
DISTRIBUTED ARCHITECTURE✳LOCAL-FIRST SYSTEMS✳WEBGL2 SHADER PIPELINES✳CONTEXTUAL LLM WORKFLOWS✳WEBASSEMBLY RUNTIMES✳ZERO-BURN COMPUTING✳
A little about me

Relentless precision.
First-principles thinking.

S.

I'm Sourav — a systems-minded product architect and engineering mentor based in Delhi, India.

I design and implement digital architectures that eliminate superfluous abstraction. Whether optimizing WebGL2 draw calls, crafting resilient multi-model failover topologies, or architecting cryptographic proof mechanisms, my standard is deterministic performance and absolute clarity.

True engineering elegance happens where algorithmic rigor intersects intuitive ergonomics. I focus on building resilient software foundations capable of handling volatile real-world demands without degradation. I also mentor engineering cohorts in modern distributed systems, WebAssembly, and browser graphics.

Core focus areas

Distributed systems · WebAssembly · browser-native storage · LLM orchestration

Active research & development

Zero-copy binary protocols · local-first state sync · GPU compute in the browser

Selected systems · 2026

High-leverage engineering,
made tangible.

Two production-grade architectures, each solving deep compute and interface constraints.

01 LernexAI
System 01S

Autonomous Technical Learning & Execution Engine

LernexAI

A distributed technical education platform architected with in-browser WebAssembly sandboxes, dynamic multi-provider LLM failover pipelines, proctored runtime verification, and cryptographic credential attestation.

Architectural Scope

Lead Systems Architect & Full-Stack Engineer

  • WASM sandboxed runtimes
  • Resilient LLM routing & failover
  • Supabase RLS & cryptographic attestation
Deep-dive system specs

LernexAI // ACTIVE RUNTIME

System Architecture
PILLARS 1-3 OF 06
01

Zero-Latency WASM Sandboxes

Browser-isolated WebAssembly runtime sandboxes for C, Python, Java, JavaScript, and SQL with deterministic memory fencing and instant execution.

02

Sub-200ms LLM Failover Router

Autonomous multi-model inference pipeline with dynamic latency profiling and seamless failover across Groq LPUs and Gemini Flash models.

03

Proctored Runtime & Anti-Cheat

Real-time session integrity tracking, fullscreen lockdown enforcement, tab-switch telemetry, and background event recording.

Page 1 of 2
Engine Stack
React 19TypeScriptWebAssemblySupabase RLSGroq LPUEdge Compute
System 02S

Browser-Native GPU Creative Workstation

Corex Quantum Studio

A browser-native creative workstation converting natural-language intent into editable bezier scene graphs, powered by hardware-accelerated WebGL2 pipelines and zero-copy Origin Private File System persistence.

Architectural Scope

Principal Architect & Graphics Engineer

  • WebGL2 fragment shaders
  • Deterministic 60 FPS scene graph
  • OPFS binary storage & ring-buffer ledger
Deep-dive system specs

Corex Quantum Studio // ACTIVE RUNTIME

System Architecture
PILLARS 1-3 OF 06
01

Natural-Language Intent AST Compiler

Four-stage deterministic planner parses design prompts into non-destructive vector hierarchies and mathematical geometry operations.

02

Real-Time 60 FPS WebGL2 Vector Stage

Hardware-accelerated GPU render loop generating dynamic bezier primitives and shaders without rasterization bottlenecks.

03

Origin Private File System (OPFS) Vault

Direct-to-disk .cxbin binary serialization with microsecond read/write persistence and zero memory leaks.

Page 1 of 2
Engine Stack
React 19TypeScriptWebGL2 / GLSLOPFS StorageZustand LedgerSIMD
Systems Topology & Engine Design

The thinking
behind the screen.

A rigorous architectural breakdown of data flows, memory isolation boundaries, and distributed state machines. Slide through dedicated ASCII blueprints below.

LERNEXAI // 01 TOPOLOGY

DIAGRAM 01 / 05
ASCII / 01 TOPOLOGYSLIDE 01 OF 05
┌────────────────────────────────────────────────────────┐
│             DISTRIBUTED LEARNER INTERFACE              │
│   WASM Code Sandbox  │  Contextual AI Stream  │ Exam   │
└───────────────────────────┬────────────────────────────┘
                            │ Non-blocking WebSocket / REST
                            ▼
┌────────────────────────────────────────────────────────┐
│           HIGH-CONCURRENCY SERVICE GATEWAY             │
│   ├─ Dynamic LLM Router (Sub-200ms TTFT Failover)      │
│   ├─ Cryptographic Signature & Payment Verifier        │
│   └─ Zero-Trust Session & State Hydration              │
└─────────────┬────────────────────────────┬─────────────┘
              │                            │
              ▼                            ▼
  ┌───────────────────────┐    ┌───────────────────────┐
  │   GROQ LPU / GEMINI   │    │  PAYMENT & EVENT BUS  │
  │  Multi-Model Pipeline │    │  Razorpay / Webhooks  │
  └───────────┬───────────┘    └───────────────────────┘
              │
              ▼
┌────────────────────────────────────────────────────────┐
│            PERSISTENCE & SECURITY FABRIC               │
│   ├─ Supabase PostgreSQL with Granular RLS Policies    │
│   ├─ Real-time Snapshot Ledger & Event Log             │
│   └─ Ed25519 Cryptographic Certificate Proofs          │
└────────────────────────────────────────────────────────┘
Distributed Learning Infrastructure

High-Concurrency Service Gateway & Core Mesh

End-to-end dataflow linking the learner surface to serverless API gateways, dynamic multi-model LLM inference failover, payment event buses, and Supabase PostgreSQL persistence.

01

WASM Learner Surface

In-browser compilation and execution sandboxes for C, Python, Java, JS, and SQL.

02

Service Gateway

Sub-200ms TTFT multi-model LLM failover, webhook validation, and zero-trust session hydration.

03

Persistence & Attestation

Supabase PostgreSQL RLS policies and Ed25519 cryptographic certificate proofs.

01 TOPOLOGY · 1 / 5Use ← → or swipe to move
Engineering principles
01

Deconstruct to first principles.

Strip away superficial assumptions. Model domain invariants, memory constraints, and runtime throughput before committing to architecture.

02

Unify system power with human clarity.

Engineer low-level systems with uncompromising efficiency while delivering ergonomic, instantaneous feedback to the end user.

03

Deterministic, continuous benchmarking.

Validate assumptions through profiling, micro-benchmarks, and fault injection to guarantee mission-critical durability.