MUTHARASU A a-mutharasu

🧑‍💻 About Me

class Mutharasu:
    
    role       = "Senior Software Engineer"
    company    = "10xScale.ai"
    location   = "Hyderabad, India"
    languages  = ["Python", "Java", "JavaScript", "SQL","Typescript","C","Rust","c++"]
    
    architect_of = {
        "Identity Provider": "Centralized OAuth 2.0 + OIDC → 5 products",
        "ETL Pipelines":     "Airflow + PySpark → 200K+ records",
        "Search Infra":      "Typesense + pgvector → semantic search",
        "Web Scrapers":      "Redis/Celery → distributed workers"
    }
    
    open_source = "Typesense #2638 — Raft consensus thread starvation"
    taught      = "Java → 200+ students"
    superpower  = "I build the platform others build on."

🏗️ What I've Built

🔐 Centralized Identity Provider

OAuth 2.0 + OIDC compliant IdP for 5 products

JWT with EdDSA signing · PKCE for public clients
Refresh token families with cascade revocation
Per-product token isolation & subscription-based auth
29 endpoints · Auth, Session, OIDC, Audit routers
FastAPI + Tortoise ORM + asyncpg + Redis + Taskiq

📊 ETL Pipeline Architecture

End-to-end data processing from scratch

Airflow orchestration + PySpark transforms
200K+ recruitment records processed
Idempotent design — safe re-execution guaranteed
PostgreSQL + pgvector for semantic search
Schema-inconsistency handling across sources

🕷️ Distributed Web Scraping

High-throughput scraping infrastructure

Source-specific worker pools (LinkedIn, Naukri)
Redis/Celery task distribution
Planned Kubernetes + KEDA autoscaling
Rate limiting & anti-detection strategies

🔀 Open Source — Typesense (26K+ ⭐, C++) Merged contributor · 3 PRs shipped

#3010 — Non-deterministic sort ordering merged

3 identical requests → 3 different orderings, no error
Cursor vector shared across sort clauses read past its allocation
Fixed with per-clause cursor keying + regression test

#3011 — mode: sum for _eval merged

Additive ranking needed 2ⁿ - 1 rules → now needs n
4 signals: 15 rules → 4
Parser disambiguation vs existing filter syntax; __int128 overflow accumulation

#2638 — Raft thread starvation reported and fixed

Write queue blocking /health & /search during 228K bulk inserts
Proposed decoupled health endpoints + backpressure

🔐 Architecture — Identity Provider

┌──────────────────────────────────────────────────────────────────┐
│                   CENTRALIZED IDENTITY PROVIDER                  │
├──────────────────────────────────────────────────────────────────┤
│                                                                  │
│  ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐     │
│  │Product 1│ │Product 2│ │Product 3│ │Product 4│ │Product 5│     │
│  └────┬────┘ └────┬────┘ └────┬────┘ └────┬────┘ └────┬────┘     │
│       └───────────┴─────┬─────┴───────────┴───────────┘          │
│                         │                                        │
│                  ┌──────▼──────┐                                 │
│                  │ API Gateway │                                 │
│                  └──────┬──────┘                                 │
│       ┌─────────────────┼─────────────────┐                      │
│ ┌─────▼─────┐   ┌──────▼──────┐  ┌───────▼──────┐                │
│ │Auth Router│   │OIDC Router  │  │Audit Router  │                │
│ │Login, Reg │   │Authorize,   │  │Logs, Events  │                │
│ │PKCE       │   │Token, JWKS  │  │Sessions      │                │
│ └─────┬─────┘   └──────┬──────┘  └───────┬──────┘                │
│       └─────────────────┼─────────────────┘                      │
│                         │                                        │
│            ┌────────────▼────────────┐                           │
│            │   FastAPI + Async Core  │                           │
│            └────────────┬────────────┘                           │
│     ┌───────────┐ ┌────▼────┐ ┌──────────┐                       │
│     │PostgreSQL │ │  Redis  │ │  Taskiq  │                       │
│     │26 Tables  │ │Sessions │ │Async Jobs│                       │
│     │Snowflake  │ │Cache    │ │Email,    │                       │
│     │IDs+Aerich │ │Rate Lim │ │Webhooks  │                       │
│     └───────────┘ └─────────┘ └──────────┘                       │
│                                                                  │
│ Security: EdDSA JWT · PKCE · HttpOnly+SameSite · No Enumeration  │
│ Per-App: Token Isolation · Unique Audience · Refresh Families    │
└──────────────────────────────────────────────────────────────────┘

🛠️ Tech Stack

Languages & Backend

Frontend

Databases, Search & Messaging

Cloud, DevOps & Infra

Data & ETL

Auth & Security

ORM & Async

👨‍🏫 Teaching & Mentoring

🎓 200+ Students Taught	🎥 Content Creator	📊 Data Engineering
Core Java, OOP, Backend	Live → Recorded tutorials	Intermediate DE series

📈 GitHub Stats

💎 Open Source

Merged contributor — 3 PRs shipped into a 26K+ ⭐ C++ codebase

PR #3011 merged — Added mode: sum to _eval sort scoring. Expressing an additive ranking model previously required enumerating every combination (2ⁿ - 1 rules — 15 for 4 signals); it now needs n. Included parser disambiguation against existing filter syntax and __int128 overflow accumulation.

PR #3010 merged — Root-caused non-deterministic result ordering: three identical requests returned three different orders with no error. A cursor vector shared across sort clauses was reading past its allocation. Fixed by keying cursors per sort-field index, with a regression test.

Issue #2638 reported and fixed — Diagnosed Raft consensus thread starvation during 228K document bulk inserts. Proposed decoupled health endpoints and configurable write backpressure.

"I don't just write code — I architect systems, debug at the protocol level, and teach others to do the same."

Navigation Menu

MUTHARASU A a-mutharasu

Achievements