AI
Available · Remote / Async · IST · Immediate

Dhrubojyoti

Gangopadhyay

AI Engineer — Advanced RAG & Multi-Agent Systems
17 years enterprise B2B ops · Production-first

Get In Touch → GitHub ↗ LinkedIn ↗
LangGraph MCP Vapi Voice AI Playwright Stealth FastAPI PostgreSQL n8n Next.js 15
Production Repositories
19+
All public, all auditable, no toy projects
Enterprise Experience
17yr
ACC Ltd · Lafarge · MP Birla Group — AGM level
Voice Pipeline Latency
<250ms
Speech → structured CRM schema, Zoya Agent
Decision Latency Reduced
40%
Distribution DB architecture, enterprise scale
The Moat

Enterprise ops meets
production AI.

Not a researcher with no deployment depth. Not a developer with no domain context. Both — fused.

17yr
Enterprise B2B Operations · AGM level
40%
Decision latency reduction, distribution DB
22%
Supply gap reduction, incentive frameworks
+14%
YoY market share scaled
Supply chain + pricing → AI architecture
17 years managing supply chains, pricing frameworks, dealer rebate architecture, and RevOps leakage at national scale across India's largest industrial conglomerates. That domain depth is baked directly into every retrieval architecture and agent guardrail built.
ACC Ltd.LafargeMP Birla GroupAGM / Senior Manager
Production systems, not prototypes
19+ public repositories. Stateful agentic graphs, custom MCP integrations, stealth automation, voice pipelines — all optimized for resource-constrained 8GB hosts and scaled to production workloads. Every repo publicly auditable at dhrubo.shop.
LangGraphMCP stdio/SSEVapiPlaywright Stealth
Database-tier thinking
Real-time rebate evaluation logic programmed at the database level — not application layer. Translating multi-state commercial operations into database schemas is a skill set rarely found in AI engineers. That's the unfair advantage.
PL/pgSQLPostgreSQLSupabaseERP Schema Design
Technical Stack

Full-spectrum
AI engineering.

Agentic & Graph
LangGraph — Supervisor Routing
Parent-Doc Memory Checkpointing
LCEL / Dynamic Fallback Loops
Self-Correcting Agents
Multi-Agent Orchestration
RAG & Protocols
MCP — stdio/SSE JSON-RPC
Hybrid BM25 + Vector Search
Reciprocal Rank Fusion (RRF)
Parent-Child Chunking
ChromaDB / Context Compression
Voice & Realtime
Vapi — Conversational Voice AI
WebSockets — FastAPI & Node.js
Claude 3.5 Realtime API
Async Voice-to-JSON Orchestration
Structured Telemetry Parsing
Web Automation
Playwright Stealth
Fingerprint Evasion (fake-useragent)
Anti-Bot Bypass — Cloudflare/Imperva
HTML → JSON Schema Extraction
SQLite State Persistence
Databases & Backend
PostgreSQL — PL/pgSQL Stored Procs
Supabase / SQLite WAL
FastAPI / Docker / GCP / Render
n8n Workflow Automation
Redis / Google Sheets API
Frontend & LLMs
Next.js 15 / TypeScript / React
Tailwind CSS
OpenAI GPT-4o / Gemini Embeddings
Anthropic Claude 3.5
Groq Llama 3
Selected Production Projects

Built. Shipped.
Auditable.

All repos are public. No demos, no mockups — real production systems with real architecture decisions.

01
NEXUS Enterprise AI Operations Platform
True Model Context Protocol abstraction layer where every capability — RAG, GitHub, Filesystem, Browser, Voice, Postgres — is a discoverable, auditable, executable tool via stdio/SSE JSON-RPC.
PythonLangGraphMCPNext.js 15DockerGemini Embeddings
View Repository ↗
Architecture
Stateful LangGraph supervisor → MCP tool registry → dynamic routing → parent-doc memory checkpointing → audit logs
Key challenge
Every capability discoverable + auditable, prevent infinite agent cycles, 8GB RAM deployment
Constraint
Gemini embeddings + lazy-loaded Playwright for resource-constrained hosts
02
Zoya Voice CRM Integration Agent
Orchestrated Vapi to manage WebRTC and SIP telephony paths, bypassing raw audio socket streams. Parses speech → outputs structured CRM schemas at sub-250ms latency under concurrent high-throughput load.
VapiWebSocketsClaude 3.5 RealtimeFastAPISupabase
View Repository ↗
Latency
<250ms speech-to-structured CRM schema — zero UI-blocking
Reliability
Supabase direct write + enterprise webhook triggers with full telemetry logging
Concurrency
Concurrent async intake handles high-throughput inbound calls without blocking
⚡ <250ms · Speech → Typed CRM Schema · Production
03
Vayu OS — High-Evasion AI Outreach Pipeline
High-evasion crawler with browser fingerprint spoofing bypassing Cloudflare/Imperva anti-bot shields. Custom SMTP handshake engine queries MX records to validate mailboxes before transmission.
PythonPlaywright StealthGroqSQLiteGoogle Sheets API
View Repository ↗
Anti-bot bypass
Playwright Stealth + fake-useragent fingerprint spoofing — Cloudflare/Imperva
Email validation
Direct SMTP MX-record handshake before transmission — zero bounce-backs
Crash resilience
SQLite transactional state manager — self-healing, restarts from last position
04
B2B Sales Intelligence RAG
Multi-agent RAG network analyzing competitor price sheets, running hybrid vector retrievals, feeding live context into negotiation systems in real time. 17 years of domain knowledge baked into retrieval architecture.
PythonLangGraphHybrid RAGFastAPIMIT License
View Repository ↗
Domain depth
17yr enterprise pricing + distribution expertise encoded in retrieval architecture
Search method
Hybrid BM25 + vector search with Reciprocal Rank Fusion
Routing
LangGraph multi-agent routing for real-time negotiation context delivery
05
Dealer Rebate Intelligence System
Real-time rebate evaluation logic programmed at the database level — not application layer. Eliminated multi-million dollar RevOps audit leakage. A decade of national distributor network management encoded in PL/pgSQL.
PL/pgSQLPostgreSQL
View Repository ↗
Architecture choice
DB-tier logic vs application layer — correctness enforced at the data source
Business outcome
Multi-million dollar RevOps audit leakage eliminated
Domain basis
10+ years managing national distributor networks across India's industrial sector
GitHub · dhruboshop

24 repos.
All live.

Every line of code is public. No toy projects. Real architecture decisions on every commit.

24
Public Repos
55+
Contributions · 12mo
Jan '25
Active Since
Python
2 days ago
EDITH
Latest build — advanced AI operations system
Updated Jun 5, 2026
Python
NEXUS-Enterprise-AI-Platform
MCP abstraction — RAG, Browser, Voice, Postgres as discoverable JSON-RPC tools
Updated last week
Python
signalstack-ai
AI signal processing and intelligence stack
Updated last week
Python
realtime-fraud-detection-agent
Real-time fraud detection using agentic reasoning and stateful graph evaluation
Updated last week
Python
Zoya-Voice-Agent
Vapi + Claude 3.5 Realtime. Speech → CRM schema at <250ms
Updated 2 weeks ago
Python · MIT
Multi-Agent-RevOps-Orchestrator
LangGraph RevOps multi-agent system — domain expertise in graph logic
Updated 2 weeks ago
Python · MIT
b2b-sales-intelligence-rag
Hybrid RAG → competitor pricing → real-time negotiation context
Updated 2 weeks ago
PL/pgSQL
loyalty_intelligence_system
DB-tier rebate evaluation logic. Eliminated multi-million RevOps leakage
Updated 2 weeks ago
TypeScript
vayu-os-frontend
AI-native CRM platform — Next.js 15 + TypeScript
Updated 3 weeks ago
View All 24 Repositories on GitHub ↗
Experience

The foundation
that separates.

Jan 2025 — Present
Self-Employed · Remote
Independent AI Engineer
  • Shipped 19+ public production repos — stateful graphs, advanced RAG networks, custom agent tracing backends, low-latency API wrappers
  • Built Zoya Voice Intake Agent: Vapi + WebRTC/SIP → structured CRM schema at <250ms latency under concurrent load
  • Engineered NEXUS — full MCP abstraction layer with discoverable, auditable tool registry via stdio/SSE JSON-RPC
  • Designed B2B RAG networks with LangGraph multi-agent routing for real-time negotiation context
  • Built high-evasion Playwright crawler bypassing Cloudflare/Imperva with SMTP MX-record validation (zero bounce-backs)
LangGraphMCPVapiFastAPIPostgreSQLPlaywright StealthSupabase
2007 — 2024
ACC Ltd · Lafarge · MP Birla Group
AGM & Senior Manager — Revenue Operations
  • Directed commercial operations, designed distribution databases — reduced decision latency by 40%
  • Architected incentive frameworks and distributor analytics — minimized supply gaps by 22%, scaled market share +14% YoY
  • Translated multi-state commercial operations into database schema rules and structured ERP technical specification blueprints
  • Built national dealer rebate systems at DB tier — eliminated multi-million dollar RevOps audit leakage
Supply ChainRevOpsPricing FrameworksDistribution DBERP Specs
2005 — 2007
University of Kalyani
MBA — First Class
  • Specialization in Operations Management & Marketing Strategy
Open To

What I'm looking for.

AI Systems Architect
Designing production agentic systems. Multi-agent orchestration, MCP integration, stateful graph design, observability-first architecture.
Applied AI Engineer
Taking AI from prototype to production. RAG systems, voice pipelines, automation infrastructure. Real deployment, not research.
AI Platform Engineer
Building the infrastructure layer for AI products. Tool registries, evaluation pipelines, memory systems, observability backends.
Founding Engineer
Early-stage AI-native products where domain depth matters. B2B, enterprise, supply chain, RevOps — 17 years of real-world context.
Remote / Async / IST (UTC+5:30)  ·  Immediate availability  ·  Open to globally asynchronous B2B collaborations  ·  All repos auditable at dhrubo.shop
Contact

Let's build
something real.

Email
gangulydhrubo@gmail.com

For detailed project discussions, enterprise AI systems, founding engineer roles, and formal business inquiries.

Send Email →
Phone / WhatsApp
+91 98740 76688

Direct line for availability discussions, project scoping, and technical conversations. IST timezone — Kolkata, India.

WhatsApp →
LinkedIn ↗ GitHub ↗ dhrubo.shop ↗ X · @aibabahq ↗ gangulydhrubo@gmail.com Kolkata, India