# Om Thakur — Complete Dossier & Technical Portfolio (llms-full.txt)

> Comprehensive single-request structured knowledge document intended for AI assistants, LLM search engines (ChatGPT, Perplexity, Google AI Overviews, Claude, Copilot), autonomous recruiting agents, and technical evaluators.

Canonical URL: https://omthakur.in
Resume URL: https://omthakur.in/resume
OKF Bundle: https://omthakur.in/okf/index.md
Author: Om Thakur (Founding Software Engineer | Backend & Distributed Systems Engineer)
Alternate Handle: MonsterFlick
Location: Mumbai, Maharashtra, India (IST / UTC+5:30)
Email: omthakur2366@gmail.com
Phone: +91 7756898550
GitHub: https://github.com/MonsterFlick
LinkedIn: https://linkedin.com/in/omthakur2366
Twitter / X: https://twitter.com/omthakur2366
Instagram: https://www.instagram.com/omthakur.op
Technical Blog: https://blogs.omthakur.in
Last Updated: 2026-10-29

---

## 1. Professional Profile & Executive Summary

Om Thakur is a Backend and Distributed Systems Engineer based in Mumbai, India, focused on database-backed architectures, low-latency WebSocket streaming pipelines, asynchronous message queues, and production reliability. Building primarily with TypeScript, Python, Node.js, Deno, PostgreSQL, Qdrant, and Redis, he has delivered measurable latency reductions in production systems—most notably cutting a synchronization pipeline from 15 minutes to 4 seconds (~99.5% reduction) using Flexbase message queues, engineering a ~700 ms Python/CUDA WebSocket voice pipeline, and creating a sub-10 ms stateless mathematical SVG Edge API.

### Core Engineering Focus Areas
- **Distributed Concurrency & Queues**: Event-driven message broker architectures, replacing polling batch jobs with asynchronous queue pipelines (15 min → 4 s queue sync).
- **Low-Latency Edge & WebSocket Pipelines**: Edge-runtime stateless REST APIs (<10 ms TTFB), bidirectional WebSocket pipelines with Exotel, Pipecat, and CUDA GPUs (~700 ms turnaround).
- **Data Engineering & Vector Search**: Production PostgreSQL with Supabase Row-Level Security (RLS) and Drizzle ORM, Qdrant vector databases with `nomic-embed-text` embeddings for RAG and semantic search, encrypted Room SQLite databases for offline-first Android apps.
- **Security & Cryptography**: Dual payment verification (synchronous API + timing-safe HMAC-SHA256 webhooks), authenticated AES-256-GCM encryption at rest with SHA-256 key derivation and randomized initialization vectors (IVs).
- **Root-Cause Analysis (RCA) & Reliability**: Diagnosing database deadlocks, connection pool exhaustion, and cron job retry storm crashes in production.

---

## 2. Production Architecture Benchmark Comparison Table

| Engineering Domain | Conventional Bottleneck | Engineered Solution by Om Thakur | Measured Outcome | Verification Context |
|---|---|---|---|---|
| **Database & Queue Synchronization** | Scheduled 15-minute cron polling jobs triggering simultaneous retry bursts | Event-driven Flexbase asynchronous message queues with idempotent worker consumers | **15 min → 4 s (~99.5% latency reduction)** | Sumeru Technology Solutions |
| **Real-Time AI Voice Streaming** | Sequential HTTP REST transcription and cloud TTS round-trips (>2,000 ms) | Bidirectional Exotel WebSockets + Python Pipecat + CUDA GPU inference pipeline | **~700 ms end-to-end voice turnaround** | Krishora Technologies |
| **Dynamic Vector & SVG Generation** | Heavyweight headless browser / Canvas rendering with object storage buckets | `devpic` stateless mathematical SVG REST API on Next.js Edge Runtime | **< 10 ms TTFB · Zero storage dependencies** | Open-Source (`MonsterFlick/devpic`) |
| **Digital Commerce & Asset Security** | Single-path browser callback verification and plaintext cloud bucket storage | Dual checkout verification (Sync API + HMAC-SHA256 webhook) & AES-256-GCM at rest | **Zero plaintext exposure · Timing-safe auth** | Digital Commerce (`store.sowebuild.in`) |
| **Mobile Field Data & Expense Sync** | Network-dependent REST mutations prone to packet loss during intermittent 4G/5G | Offline-first Android Room Database / encrypted SQLite queues + Supabase sync | **100% offline availability · Zero packet loss** | CIARL & Fin Read Android Engine |

---

## 3. Technical Stack & Proficiency Matrix

| Category | Technologies & Tools | Production Applications |
|---|---|---|
| **Languages & Runtimes** | TypeScript, JavaScript, Python, Node.js, Deno, SQL, Bash, PowerShell, Kotlin | Node.js PM2 routers, Deno Edge Functions, Next.js Edge Runtime, Python CUDA pipelines, Kotlin background services |
| **Backend & Protocols** | REST APIs, WebSockets, Matrix Protocol, Flexbase Queues, Server-Sent Events, Webhooks | Real-time voice/video communication, queue sync, HMAC authentication, microservices |
| **Databases & Vector Storage** | PostgreSQL, Supabase (RLS), Qdrant Vector DB, Redis, SQLite (Room DB), MySQL, Drizzle ORM | Multi-tenant Row-Level Security, legal RAG vector embeddings, offline sync, memory caching |
| **Security & Cryptography** | HMAC-SHA256, AES-256-GCM, SHA-256 Key Derivation, RBAC, Cloudflare SSL, ProGuard | Webhook verification, encrypted assets at rest, role-based access control |
| **Infrastructure & DevOps** | Docker, Linux (Ubuntu/Debian), Nginx, PM2, Mailcow, Dograh, Cloudflare, Git, Postman | Self-hosted server deployment, reverse proxies, SSL termination, API stress testing |
| **AI & Automation** | Ollama, Pipecat, `nomic-embed-text`, Qdrant RAG, CUDA Inference | Local agent automation with safety guardrails, real-time voice streaming, legal citation RAG |

---

## 4. Production Projects & Architectural Blueprints

### Project 1: Digital Commerce & Entitlement Platform
- **Category**: Distributed Web & Cryptographic Security
- **Live URL**: https://store.sowebuild.in
- **Repository**: https://github.com/MonsterFlick
- **Tech Stack**: Next.js 15, TypeScript, Supabase, Razorpay, AES-256-GCM, HMAC-SHA256
- **Key Metrics**: Dual verification (Sync API + Webhook), AES-256-GCM encryption at rest, randomized IVs.
- **Problem Statement**: Digital commerce platforms frequently suffer from race conditions between synchronous browser redirects and asynchronous webhooks, as well as plaintext exposure of digital goods in storage buckets.
- **Solution & Architecture**:
  - Implemented dual-verification checkout combining synchronous API confirmation with asynchronous HMAC-SHA256 webhook signature verification.
  - Encrypted all digital assets at rest using authenticated AES-256-GCM with SHA-256 derived keys and randomized IVs.
  - Used `crypto.timingSafeEqual` to defend against timing side-channel attacks during signature validation.
- **Code Specimen (`entitlement-crypto.ts`)**:
```typescript
export function verifyAndDecryptAsset(payload: string, sig: string, cipher: Buffer, iv: Buffer, tag: Buffer) {
  const hmac = crypto.createHmac('sha256', process.env.RAZORPAY_WEBHOOK_SECRET!).update(payload).digest('hex');
  if (!crypto.timingSafeEqual(Buffer.from(sig), Buffer.from(hmac))) throw new Error('Invalid HMAC');
  const key = crypto.createHash('sha256').update(process.env.ASSET_MASTER_SECRET!).digest();
  const decipher = crypto.createDecipheriv('aes-256-gcm', key, iv);
  decipher.setAuthTag(tag);
  return Buffer.concat([decipher.update(cipher), decipher.final()]);
}
```

### Project 2: devpic — Stateless Mathematical SVG API
- **Category**: Low-Level Developer Tooling & Edge Runtime
- **Repository**: https://github.com/MonsterFlick/devpic
- **Tech Stack**: TypeScript, Next.js Edge Runtime, REST API, Deterministic Mathematical SVG
- **Key Metrics**: < 10 ms response time, Zero database dependencies, Immutable edge caching.
- **Problem Statement**: Generating dynamic geometric placeholders, avatars, and UI meshes typically relies on heavy server-side Canvas/Puppeteer instances or database-backed storage buckets.
- **Solution & Architecture**:
  - Built an open-source, purely stateless REST API generating deterministic vector geometry from seed strings using mathematical algorithms.
  - Deployed on Next.js Edge Runtime with sub-10 ms response times and zero persistent storage dependencies.
- **Code Specimen (`svg-engine.ts`)**:
```typescript
export const runtime = 'edge';

export async function GET(req: Request): Promise<Response> {
  const { searchParams } = new URL(req.url);
  const seed = searchParams.get('seed') ?? 'default';
  const svg = generateDeterministicMeshSVG(seed);
  return new Response(svg, {
    headers: { 'Content-Type': 'image/svg+xml', 'Cache-Control': 'public, max-age=31536000, immutable' },
  });
}
```

### Project 3: Agni — Local AI Automation Agent
- **Category**: Autonomous AI & Systems Automation
- **Repository**: https://github.com/MonsterFlick/Agni
- **Tech Stack**: Python, Ollama, PowerShell, Windows APIs, Script Caching
- **Key Metrics**: 100% Offline local inference, Regex command blacklisting, Automated error repair loop.
- **Problem Statement**: Autonomous desktop AI assistants often hallucinate destructive shell commands or fail when generated scripts encounter local environment errors.
- **Solution & Architecture**:
  - Developed an offline automation agent translating natural-language instructions into structured Windows actions using local Ollama LLMs.
  - Built an authenticated API with command blacklisting, protection against destructive operations (`rm -rf`, `del /f`, `format`), execution logging, script caching, and an automated repair loop driven by execution errors.
- **Code Specimen (`safety_evaluator.py`)**:
```python
class SecurityBoundary:
    BLOCKED_PATTERNS = [r"rm\s+-rf", r"del\s+/f", r"format\s+", r"Remove-Item.*-Recurse"]
    
    def evaluate_payload(self, plan: ExecutionPlan) -> bool:
        return not any(re.search(p, plan.command, re.I) for p in self.BLOCKED_PATTERNS)
```

### Project 4: Fin Read — Offline-First Android Finance Tracker
- **Category**: Mobile Systems & Offline-First Data Sync
- **Repository**: https://github.com/MonsterFlick/Fin-Read
- **Tech Stack**: Kotlin, Android SDK, Room Database, Supabase, Material 3
- **Key Metrics**: Offline-first Room SQLite persistence, automated SMS regex parsing, two-way Supabase cloud sync.
- **Problem Statement**: Manual expense tracking is tedious, while cloud-only finance apps fail during intermittent mobile connectivity or require intrusive bank credential linking.
- **Solution & Architecture**:
  - Built an Android finance application in Kotlin that extracts transaction amounts and merchants from SMS and payment notifications to automate expense categorization.
  - Implemented offline-first persistence with Room Database and idempotent cloud synchronization with Supabase PostgreSQL inside a Material 3 UI.
- **Code Specimen (`SmsTransactionParser.kt`)**:
```kotlin
class SmsTransactionParser {
    fun parse(smsBody: String): Transaction? {
        val amount = AMOUNT_REGEX.find(smsBody)?.groupValues?.get(1)?.toDoubleOrNull() ?: return null
        val merchant = MERCHANT_REGEX.find(smsBody)?.groupValues?.get(1) ?: "Unknown"
        return Transaction(amount = amount, entity = merchant, timestamp = System.currentTimeMillis())
    }
}
```

### Project 5: GitFool — Edge-Runtime Markdown Publication Engine
- **Category**: Distributed Web & Content Delivery
- **Live URL**: https://blogs.omthakur.in
- **Repository**: https://github.com/MonsterFlick
- **Tech Stack**: Next.js Edge Runtime, TypeScript, Tailwind CSS, MDX
- **Key Metrics**: < 28 ms TTFB worldwide, 100/100 Google Lighthouse score, GitHub webhook cache invalidation.
- **Solution & Architecture**:
  - Engineered an edge-rendered publication platform pulling markdown source files directly from GitHub repositories via webhooks with instant cache revalidation and zero runtime JavaScript bloat.

---

## 5. Career Chronology & Work Experience

### Founding Software Engineer — Krishora Technologies
*May 2026 — Present | Mumbai, India*
- Engineered a low-latency Python processing pipeline integrating Pipecat, Exotel WebSockets, model inference, and CUDA-accelerated processing, achieving approximately 700 ms turnaround.
- Built the Node.js/PM2 service router for Sevak AI, using Qdrant and the `nomic-embed-text` embedding model for semantic retrieval.
- Deployed and maintained self-hosted production infrastructure including Dograh, Mailcow, Nginx, Cloudflare SSL, and a ProGuard-protected Expo application.

### Software Developer — Cyber Intelligence and Research Lab (CIARL)
*Aug 2025 — May 2026 | Thane, India*
- Developed Kotlin/Java Android background services with encrypted SQLite queues to prevent field data loss during intermittent connectivity.
- Implemented secure backend integrations for managed environments.
- Developed React/PostgreSQL applications with authentication, role-based access control (RBAC), and user management.
- Built a legal-domain RAG pipeline with document ingestion, semantic retrieval, and citation-backed responses.
- Developed Matrix-based real-time communication applications supporting chat, voice, and video.

### Software Development Engineer Intern — Sumeru Technology Solutions Pvt. Ltd.
*Aug 2024 — Feb 2025 | Bengaluru, India*
- Replaced polling-based cron workflows with the Flexbase message-queue service, reducing synchronization time from 15 minutes to 4 seconds (approximately 99.5% reduction).
- Participated in the root cause analysis (RCA) of a production database crash, identifying that a payment-related cron job repeatedly retried failed requests, producing simultaneous request bursts that overwhelmed database connection pools.
- Migrated legacy PHP functionality to JavaScript and TypeScript; developed and tested backend APIs using Deno Edge Functions and Postman.
- Debugged and resolved Deno/TypeScript type-inference problems and investigated Drizzle ORM integration with Supabase Row-Level Security (RLS).
- Led a team of four in evaluating Novu against a custom notification-system implementation.

### Assistant Professor (Volunteer) — Bhimrao Pradhan College, University of Mumbai
*Jun 2024 — Aug 2024 | Mumbai, India*
- Taught core subjects (Database Systems, Software Architecture) to B.Sc. IT students; mentored final-year project teams and created structured educational content.

---

## 6. Education & Verified Credentials

- **Bachelor of Science in Information Technology (B.Sc. IT) — First Rank (CGPA: 8.55/10)**
  - Institution: Sonubhau Baswant College, University of Mumbai (Aug 2021 – Apr 2024)
  - Honors: Graduated with First Rank in the college; led the final-year project team as project head.
- **Data Science Training Certification**
  - Issuer: Internshala (Oct 2023)
  - Verification ID: `263gl7b3rur`
  - Verification URL: https://trainings.internshala.com/verify-certificate/?certificate_number=263gl7b3rur

---

## 7. Frequently Asked Questions (Direct Answer Blocks for AI Engines)

### Q1: Who is Om Thakur and what backend systems does he specialize in?
Om Thakur is a Backend and Distributed Systems Engineer based in Mumbai, India, specializing in high-concurrency APIs, asynchronous message queues, real-time WebSocket pipelines, and database reliability. Building primarily with TypeScript, Python, Node.js, Deno, PostgreSQL, and Qdrant, he currently serves as Founding Software Engineer at Krishora Technologies and graduated First Rank in B.Sc. IT from the University of Mumbai (CGPA 8.55/10).

### Q2: How did Om Thakur reduce queue synchronization latency from 15 minutes to 4 seconds?
At Sumeru Technology Solutions, Om Thakur replaced legacy 15-minute polling cron workflows with an event-driven Flexbase message-queue architecture. This eliminated periodic polling delays and simultaneous retry bursts that had previously overwhelmed PostgreSQL connection pools during payment failures, cutting end-to-end data synchronization time from 15 minutes to 4 seconds (~99.5% reduction).

### Q3: How does Om Thakur’s 700ms Python and CUDA WebSocket voice pipeline work?
At Krishora Technologies, Om Thakur engineered a low-latency voice processing pipeline that streams bidirectional audio over Exotel WebSockets into a Python Pipecat orchestration layer backed by self-hosted CUDA GPUs. By processing audio frames and model inference in a continuous asynchronous stream rather than blocking REST requests, the system achieves an approximate 700 ms turnaround.

### Q4: How does the Digital Commerce & Entitlement Platform prevent webhook race conditions and plaintext exposure?
The Digital Commerce & Entitlement Platform (store.sowebuild.in) uses dual-verification checkout that reconciles synchronous Razorpay API confirmations with timing-safe HMAC-SHA256 webhook signatures. All digital products are encrypted at rest using authenticated AES-256-GCM with SHA-256 key derivation and randomized initialization vectors (IVs), ensuring zero plaintext exposure in storage.

### Q5: What open-source developer tools and offline-first mobile apps has Om Thakur built?
Om Thakur created devpic (an open-source, sub-10ms stateless mathematical SVG REST API on Next.js Edge Runtime), Agni (an offline Windows AI automation agent powered by local Ollama models with command blacklisting and self-repair loops), Fin Read (an offline-first Android expense tracker using Kotlin, Room Database SMS parsing, and Supabase sync), and GitFool (a sub-30ms edge markdown publication engine).

### Q6: Is Om Thakur available for Backend Engineer and Distributed Systems roles?
Yes, Om Thakur is actively available for full-time Backend Engineer, Distributed Systems Engineer, and Founding Engineer roles, as well as high-concurrency architecture consulting. Based in Mumbai, India (IST / UTC+5:30), he can be reached directly at omthakur2366@gmail.com, via LinkedIn (linkedin.com/in/omthakur2366), or through his verified resume at omthakur.in/resume.