IT Manager · AI Transformation Lead · Enterprise Architect
Building the AI-native enterprise from the inside.
7+ years building enterprise systems. Obsessed with one mission: the AI-native enterprise.
About Me
The person behind the systems.
"Data doesn't just inform decisions — it prevents expensive mistakes."— My approach to every project
I'm an IT Manager and AI transformation lead with 7+ years of enterprise systems experience — and a singular obsession: what does the AI-native enterprise actually look like, and how do you build one from the inside?
My career started in the trenches: writing ERP modules, debugging marketplace integrations at 2 AM, and learning that elegant code means nothing if the business can't use it. That foundation is what makes the AI work stick — I don't just deploy agents, I rebuild how organizations operate.
Data over assumptions.
When I inherited a critically unstable production system, I didn't rush to fix it — I analyzed 65 days of logs first. That data shaped a 12-month roadmap now delivering 329% ROI. Five months in: a system that runs clean.
Ship fast, scale smart.
4 production web apps in 3 months. 101+ agents in 5 months. 19 Docker services with 99.7% uptime. I architect for resilience but ship iteratively — quick wins in Q1, compounding transformation across Q2–Q4.
AI as infrastructure, not a feature.
Most teams treat AI as a tool you bolt on. I build it into the operating system — agents with memory, MCP servers as nervous system, RAG as institutional knowledge. The result: systems that get smarter over time.
What I Do
Five capability pillars.
The domains I operate across — from code to boardroom.
Enterprise Architecture & ERP Transformation
I design and implement multi-company ERP ecosystems using Odoo v17, integrating manufacturing, rental operations, finance, inventory, and CRM into unified platforms.
AI & Agentic Systems Engineering
I build production AI systems — not prototypes. From deploying 101+ specialized AI agents using structured skill decomposition, to engineering enterprise RAG systems with 2.55M+ vectors across Qdrant, pgvector, and Redis.
Data Infrastructure & Performance Engineering
I diagnose and fix the performance problems everyone else works around. Database queries running in minutes get reduced to seconds. Systems drowning in errors get stabilized.
Integration Architecture & Automation
Enterprise systems rarely exist in isolation. I design the connective tissue — webhook-driven integrations, queue-based processing, API orchestration — that makes disparate systems function as one.
Strategic IT Leadership & Team Building
I lead technology teams and translate business strategy into technical roadmaps — and technical reality into language the business understands.
Featured Projects
Enterprise transformation in action.
Real problems, real solutions, real impact. Click any card to explore the full story.
Enterprise Database Performance Overhaul
Inherited a production database environment where critical business queries took over 4 minutes to execute. Users had normalized waiting, and workarounds were embedded into daily operations. Nobody had diagnosed the root cause — they'd just accepted it.
Approach
Conducted forensic analysis of the entire database layer. Mapped query execution plans, identified missing indexes on high-traffic tables, and designed a systematic indexing strategy targeting the highest-impact operations first.
Impact
- Primary lookup queries: 96.2% performance improvement (minutes reduced to seconds)
- Documented 892% ROI on the optimization initiative
- Eliminated user workarounds costing hours of productivity daily
- Created performance monitoring baseline for ongoing health tracking
AI Operating System for the Enterprise
Managing enterprise-scale technology transformation with a lean team meant development velocity was the bottleneck. Traditional approaches — more developers, longer timelines — weren't viable. I needed to fundamentally change how the entire organization operated.
Approach
Built a full AI operating system: 101+ specialized agents across 9 departments, 10 MCP servers with 180+ tool functions connecting agents to SAP B1, Odoo, Jira, GitHub, and Loki. 14-file context engineering system with model routing (Haiku/Sonnet/Opus) reduced LLM costs by 70-85%.
Impact
- Development throughput more than tripled through AI-augmented workflows
- Production system transformed from critically unstable to near-zero errors
- System health transformed from degraded to high-performing in 5 months
- 329% documented ROI on full transformation initiative
- 4 production web apps shipped in 3 months
Enterprise RAG Knowledge System
Organizational knowledge was scattered across databases, documents, logs, and people's heads. I built a 3-layer RAG architecture that gives the entire organization a queryable, intelligent memory.
Approach
3-layer RAG: Qdrant (2.55M+ vectors, 5 collections) for semantic search, pgvector with HNSW indexing (150ms raw → 22ms cached via Redis at 98% hit rate), and Gemini Vision for equipment image and safety document extraction at 92-95% accuracy.
Impact
- 2.55M+ vectors indexed across equipment, safety, operations, and product knowledge
- 22ms retrieval latency achieved via Redis caching (98% hit rate)
- 92-95% extraction accuracy on visual documents via Gemini Vision
- Enabled natural-language querying of enterprise data across 9 departments
Enterprise AI Platform Suite
While transforming the AI infrastructure, I also needed to ship real software. In 3 months: a project intelligence platform, a customer enrichment engine, an AI safety inspector, and a market intelligence dashboard — all in production, all AI-powered.
Apps Shipped
- Radar Y — Project intelligence platform. 2,529 construction projects tracked, 72,712 AI-scored customer-project matches, 22 pages, 93 tests. Live at radar.ygroup.cloud
- Enrichment Pipeline — Customer intelligence engine. 511 customers × 127 fields = 62,619 data points. 27 AI agents for lead scoring, churn prediction, ghost-phase detection.
- Safety Inspector — Gemini Vision AI safety analysis. Equipment image recognition, compliance checking, 92-95% extraction accuracy.
- Market Platform — RFM segmentation, 6,002 leads in pipeline, 1,897 news intelligence signals from 19 sources, 6,612 stakeholders mapped.
Stack & Scale
- 49 UI pages, 115 API routes, 105+ database tables across all apps
- 537+ automated tests
- 19 Docker production services, 99.7% uptime
- Full Grafana + Prometheus + Loki observability
Impact
Impact by the numbers.
Measurable outcomes from enterprise-scale technology transformation.
Career
The journey so far.
From writing ERP modules to building a 101-agent AI operating system.
IT Manager & AI Transformation Lead
Y Equipment Services · Y Group (YES · YAM · Y-Mesh)
What I Inherited
A fragmented enterprise: critically unstable production systems, poor compliance rates, dozens of manual hours consumed by repetitive reports, virtually no documentation. No AI. No automation.
What I Built (5 months)
- 101+ AI agents across 9 departments — sales, ops, finance, manufacturing, safety, HR, marketing, IT, CEO
- 10 MCP servers, 180+ tool functions — agents connected to SAP B1, Odoo, Jira, GitHub, Loki, Beta (152 endpoints)
- 4 production web apps — Radar Y (project intelligence), Enrichment Pipeline, Safety Inspector, Market Platform
- 3-layer RAG system — Qdrant (2.55M+ vectors) + pgvector (HNSW, 22ms) + Gemini Vision (92-95% accuracy)
- 160+ automation workflows — 21 cron jobs, 20 job queues, 78 pipeline scripts, 3 launchd daemons
- 6 custom Odoo v17 modules — 75 security groups, 240+ ACLs, autonomous SAP/Beta/Exotel pipelines
- 19 Docker production services — full Grafana/Prometheus/Loki observability, 99.7% uptime
Results
- Production system transformed from critically unstable to near-zero errors
- Overall system health transformed from degraded to high-performing
- Development throughput more than tripled through AI-augmented workflows
- SOPs: 3 → 73 in 5 months
- 329% documented ROI on full transformation initiative
- 120+ hours/month freed through hyperautomation
Associate Project Manager
InnoAge Technologies PVT. LTD.
Where the AI Journey Began
This is where everything shifted. Managing a team of 10 engineers wasn't the challenge — finding a way to make that team perform like 50 was. AI became the answer.
The AI Experiments That Changed Everything
- All-Framework Agentic Kit — built a custom agentic SDLC kit for Odoo development. Result: 10X faster MVP delivery. Teams that took months shipped in weeks.
- Cursor IDE rollout — trained the entire engineering team on AI-first development. Dev velocity multiplied 5X. Not measured in vibes — measured in delivered features.
- n8n automation — built real-time task generation pipelines from meeting transcripts and emails. Tasks went from "discussed" to "assigned in Jira" automatically.
Projects Delivered
- Surya India MRP — manufacturing ERP for one of India's largest carpet manufacturers. Full production planning, WIP tracking, quality control.
- Chandrawoollens ERP — end-to-end ERP for yarn spinning and dyeing operations. Deployed Apr 2025.
- UK Warehouse On-Site — visited UK warehouse Jan 2025. Diagnosed and resolved live WMS issues for 100K+ rug unit operations.
- Odoo v15 → v17 Migration — led full platform migration Feb 2025 across complex multi-module environment.
Lead Product Engineer
InnoAge Technologies PVT. LTD.
The Leadership Shift
Moved from building to leading. Responsible for a cross-functional team of 5-6 engineers across the full delivery lifecycle — from architecture decisions to deployment.
What I Shipped
- Surya Europe ERP — B2C/B2B ERP platform across EU markets with exports from Turkey and India. Complex multi-currency, multi-warehouse, multi-language operation.
- UK Warehouse WMS — Odoo Barcode WMS tracking 100,000+ rug units across a UK warehouse. Real-time scanning, putaway rules, cycle counts.
- EU Marketplace Integrations — connected 15+ European platforms (Amazon EU, Wayfair, Mirakl, CDiscount, and more) into a unified inventory and order management system.
- CI/CD Pipelines — established automated deployment workflows for enterprise Odoo, cutting release risk and time-to-production.
Leadership
- Established code review culture and engineering standards across the team
- Mentored junior engineers on ERP architecture and marketplace integration patterns
Senior Software Engineer
InnoAge Technologies PVT. LTD.
Deep ERP + EU Scale
First exposure to enterprise ERP at European scale — multi-country, multi-currency, multi-warehouse operations where edge cases are the norm.
What I Built
- Marketplace hub architecture — designed the core integration layer connecting 15+ EU sales channels to a unified Odoo inventory. Every order from every platform, one source of truth.
- Custom WMS — real-time stock synchronization across multiple warehouse locations, automated stock reservation, and fulfillment routing logic.
- Carrier management system — integrated shipping carriers (DHL, DPD, Hermes) with automated label generation, tracking updates, and returns processing.
- EU compliance modules — VAT handling across multiple EU jurisdictions, OSS scheme integration, fiscal position automation.
Senior Software Engineer
TraceNCode Technologies
The Consulting Chapter
Moved from product to consulting — now the client's problems were my problems to diagnose and solve, often with legacy codebases and tight timelines.
What I Built
- Cross-version ERP migrations — migrated clients from older Odoo versions to newer ones, preserving data integrity, custom logic, and business continuity.
- Client-specific customizations — rapid-cycle development across diverse industries: retail, distribution, manufacturing. Each client a different puzzle.
- Structured client engagement — developed requirement gathering and scoping processes that reduced mid-project scope creep significantly.
What This Taught Me
Consulting sharpens you fast. No time to over-engineer. No room for assumptions. You learn to ask the right questions before writing a single line of code.
Software Engineer
Webkul Software PVT. LTD.
The Foundation Years
Fresh out of B.Tech with a 9.0 GPA. Joined one of India's leading Odoo product companies and immediately started shipping — modules, integrations, connectors. The volume here was the education.
What I Built
- 40+ marketplace connector modules — Amazon EU, Wayfair, Mirakl, Google Shopping, Facebook Shop, CDiscount, and more. Each one a different API, different data model, different edge cases.
- 50+ custom ERP modifications — across manufacturing, distribution, retail, and e-commerce clients. Every modification a new lesson in how businesses actually work vs how they think they work.
- Cloud storage integrations — S3, Google Cloud, Dropbox connected to Odoo document management. Files that live where the business needs them.
- Frontend theme systems — custom Odoo website themes and e-commerce storefronts. The user-facing layer of enterprise operations.
- Odoo core modules — deep work across Inventory, Sales, Purchase, Manufacturing, Accounting, HR. Built the mental model of ERP that everything since has run on.
What These Years Built in Me
Volume builds instinct. 40+ integrations means you stop asking "how do I connect these?" and start asking "why are they disconnected in the first place?" That question is what eventually led to AI.
Skills
The journey to AI-native.
Not a list of tools. A journey — from first line of code to building an AI operating system.
Started by chaining a few API calls. Then added tools. Then memory. Then coordination between agents. Now: 101 specialized agents running across 9 departments — each with a defined role, tools, and memory — powering a live enterprise AI operating system.
Started by writing "summarize this." Hit token limits. Learned few-shot patterns, system prompts, structured outputs. Built a 14-file context routing system — the AI now loads exactly what it needs, when it needs it. Renamed the skill: it's context engineering now.
Realized agents are only as powerful as their reach. Built 10 MCP servers from scratch — now agents can read SAP invoices, query Odoo leads, analyze Loki logs, file Jira tickets, and push GitHub PRs autonomously. The nervous system of the AI enterprise.
First tried stuffing everything into the context window. Hit walls. Learned vector embeddings, semantic search, hybrid retrieval. Now 2.55M+ vectors across Qdrant and pgvector give the enterprise a searchable long-term memory — answers in 22ms.
Deployed one agent. Then ten. Then realized AI at scale is a reliability engineering problem. Built caching strategies, model routing, budget caps, memory persistence, structured JSON outputs — the ops layer that makes 101 agents run reliably every single day.
Started with text. Then discovered equipment manuals, inspection photos, and compliance documents needed understanding too. Integrated Gemini Vision for equipment analysis and safety document extraction — 92-95% accuracy running in production.
The language everything runs on. Started with ERP modules. Moved to data pipelines, automation scripts, AI agent orchestration. Every layer of the AI operating system speaks Python.
Started writing queries. Then optimizing them. Then diagnosing production degradation through execution plans. Then adding pgvector with HNSW indexing for semantic search. A database became an intelligence layer.
Frontend first. Then N8N custom nodes. Then full-stack Next.js with Drizzle ORM and Supabase. Shipped 4 production web apps — project intelligence, customer enrichment, safety inspection, market analytics. The web layer of the AI stack.
Hundreds of Odoo views, access control files, data declarations in XML. Interactive SOPs and dashboards in HTML/CSS. The declarative and visual layers of enterprise systems.
Started writing simple modules at Webkul in 2019. Then complex integrations, version migrations, multi-company architecture. Now building ERP as the data backbone of an AI enterprise — autonomous pipelines feeding intelligence in and out.
Production orders, WIP tracking, BOM management, real-time stock sync across warehouses. Learned ERP from the factory floor up — where the real complexity lives.
Built the bridge between SAP Business One and the AI stack. OData APIs, invoice reconciliation, autonomous daily sync — SAP data now flows directly into AI enrichment pipelines without human intervention.
Multi-currency AR management, tax compliance, automated invoice chasing. Then went further: built AI agents that handle AR aging analysis and follow-up autonomously — hours of manual work, automated.
Built 40+ marketplace connector modules across Amazon, Wayfair, Mirakl, Google Shop, and 11 more EU platforms. The foundation for understanding enterprise integration at scale — every system speaks a different language.
Started running one container. Then learned Compose. Then orchestrated 19 production services — vector stores, AI servers, databases, caching, monitoring — all in a single stack running at 99.7% uptime.
Started with manual log-checking. Then built dashboards. Then real-time alerting. Now 7 Grafana dashboards monitor 19 services, with 50M+ log entries in Loki — the enterprise has full visibility into itself.
From SSH basics to managing production VPS infrastructure. Process monitoring, service management, log analysis. The ground on which the entire AI stack runs.
Reverse proxy, automatic SSL, traffic routing. Traefik handles certificate management for every production app. The gateway layer that connects the world to the AI platform.
Version control from day one. PR workflows, branch strategies, CI/CD pipelines. Then went further: AI agents now perform code review automatically — cutting review time dramatically.
Started with one webhook trigger. Then built workflows. Then chained them across systems. Now 160+ automated pipelines run across 9 departments — from lead creation to AR chasing to compliance. 120+ hours/month freed.
Connected SAP B1, Odoo, Beta (152 endpoints), Exotel, Brevo, Jira, GitHub, and more. Enterprise integration is the art of making fragmented systems speak a common language — and then letting AI agents read it all.
Built 20 database-driven job queues with priority, retry, and dead-letter handling — the reliable async layer that runs AI enrichment pipelines at scale without losing a single record.
Started using Redis for session storage. Then discovered caching as a first-class performance strategy. Now Redis sits between every RAG query and the vector database — 98% cache hit rate, 22ms retrieval instead of 150ms.
Integrated cloud telephony into CRM — click-to-call, IVR routing, DND/NDNC compliance, campaign automation. The next step: AI voice agents routing calls intelligently based on customer context.
Built a 12-month AI transformation roadmap from scratch — assessed the landscape, identified quick wins, designed the long-term architecture, scoped investment vs return. Now executing Q2 with documented 329% ROI.
The framework behind 101 agents. Business-focused Multi-Agent Design — every agent has a defined role, tools, and success metrics tied to business outcomes. AI that's built for results, not demos.
Led teams of 10+ engineers across the full development lifecycle. Capability mapping, AI-first dev training, structured code review culture. Turned a reactive team into a proactive, AI-augmented engineering organization.
Not just strategy — execution. 4 production web apps in 3 months, 537+ automated tests, full observability from day one. Requirements → architecture → build → live. Repeat.
Sprint planning, Jira administration, retrospectives, velocity tracking. Then layered AI on top — agents that generate standup summaries, track blockers, and flag delivery risks before they become problems.
Contact
Let's connect.
I'm open to conversations about enterprise technology leadership, AI systems architecture, ERP transformation strategy, and the messy reality of making large-scale systems work.