I BUILD AI AGENTS. I RUN PROD.
Hey, I'm Priyansh Gupta — Senior Agentic AI & GenAI Engineer with 5+ years delivering enterprise AI on Azure and AWS. I build multi-agent systems, production RAG pipelines, and MCP integrations — production-ready GenAI platforms trusted by Fortune 500 clients.
ANYONE CAN WEAR THE MASK. I SHIP THE AGENTS.
For 5+ years I've built and run cloud infrastructure and backend systems — Kubernetes clusters, GitOps pipelines, zero-trust service meshes — for Fortune 500 enterprises and global clients. Then GenAI arrived, and I took the leap of faith: the same engineering discipline could make autonomous agents safe enough for production.
Today I'm a Senior Agentic AI Engineer & GenAI Lead based in Dubai (JVC), designing multi-agent platforms, enterprise RAG systems, and MCP integrations — with guardrails, HITL approval flows, and full auditability. Certified Azure AI Engineer (AI-102), CKA, CKAD, and AZ-400.
"Agents you can't audit are agents you can't trust."
M.Sc. Artificial Intelligence
Heriot-Watt University · Dubai, UAE
IN PROGRESSB.Tech. Computer Science
UPES · 2017 – 2021
CLOUD & AI/MLFROM PIPELINES TO PRODUCTION GENAI
Senior DevOps & AI Platform Engineer
Deployed multimodal LLMs on Azure AI Foundry with inference auto-scaling and Content Safety guardrails. Architected enterprise RAG pipelines with vector + semantic hybrid indexing in Azure AI Search, delivered MCP integration for Azure Boards for live agentic tool-use, and led GenAI platform modernisation for FMGlobal with zero-downtime migration. Standardised CI/CD governance across 20+ microservices with JFrog supply-chain gates.
Senior GenAI / MLOps / Agentic AI Consultant
Architected end-to-end RAG pipelines with LangChain, Pinecone, and ChromaDB — semantic retrieval with top-k reranking for production SaaS clients. Delivered agentic AI solutions embedding Azure OpenAI, Claude, and GPT-4o with tool-use, function-calling, and autonomous decision-making. Deployed MLOps infrastructure on Azure ML and AWS SageMaker with n8n-orchestrated pipelines, model monitoring, and drift detection.
DevOps Engineer
Engineered Helm charts for AKS onboarding with rollback-enabled pipelines and Kyverno admission control for security baselines, quotas, and image signing. Configured Istio with mTLS zero-trust, FluxCD GitOps state management, and canary + A/B traffic routing. Enforced least-privilege secrets access via Kubernetes Secrets Store CSI and Azure Key Vault.
Software Engineer
Built reusable Terraform and Bicep modules for Azure VNets, Key Vaults, AKS, and Application Gateways; authored Azure DevOps YAML CI/CD templates for build-test-deploy workflows across cloud-native services.
AGENTS WITH GUARDRAILS, PLATFORMS THAT SCALE
Mach2
Successor to Mach1 — a multi-agent ReAct harness instrumented end-to-end with self-hosted Langfuse, tracing every thought, tool call, and LLM span. An LLM-as-a-Judge (Grok) scores each response on hallucination and quality, while a five-critic pipeline (Fact, Gap, Logic, Efficiency, Sign-off) emits a severity-tagged critique ledger that gates synthesis and triggers targeted re-research — plus a nested Research Agent fanning out scoped sub-agents over shared ChromaDB memory, streamed live in an Electron + React desktop UI.
Mach1
A local-first agent on LangChain + LangGraph (StateGraph) running open-source LLMs through Ollama — a confidence-scored triage router answers 65%+ of requests locally and escalates only hard prompts to Claude via MCP: fast, private, zero-cost. Full multi-tool loop (file ops, shell, Python REPL, web tools) with episodic + semantic ChromaDB memory, top-k RAG recall, a periodic summarizer agent, and a three-mode React + TypeScript UI with live confidence and routing tags.
Ornith 1.0
A desktop coding-agent harness (Next.js + Electron, React + TypeScript) for the self-scaffolding ornith:9b model on local Ollama — each session binds its own pluggable Python agent harness that plans tasks and streams native tool-calls (file edits, reads, shell runs). An Inspector / Thinking / Critiques workspace surfaces ReAct reasoning and an L2–L7 critic ledger per response, with a harness registry and a full Playwright e2e suite.
BlastR
M.Sc. dissertation: parses Terraform/IaC into NetworkX attack graphs and propagates "blast radius" across cloud layers, pairing a LangGraph reasoning agent with FAISS + Sentence-Transformers RAG to explain and prioritise security misconfigurations. Fine-tuned Qwen2.5-Coder-1.5B with LoRA/QLoRA on Checkov/Trivy-labelled Terraform, benchmarked against a zero-shot baseline and tracked in LangSmith — shipped as a Typer CLI, FastAPI service, and React dashboard.
CrewSpace
Built from scratch: a platform where a CEO agent autonomously decomposes goals into domain subtasks, delegates across specialist AI agents, proposes new hires when needed, and coordinates via multi-agent communication and approval workflows. Ships with persistent per-agent memory, agent-to-agent and group chat, shared workspaces with file permissions, integrated terminal ops, real-time task boards, founder dashboards, and an interactive 3D office.
FMGlobal — GenAI Modernisation
Led the modernisation of FMGlobal's GenAI platform — zero-downtime migration of critical insurance workloads to secure Azure AI infrastructure. Multimodal LLM endpoints on Azure AI Foundry with auto-scaling and Content Safety guardrails, enterprise RAG with hybrid vector + semantic indexing in Azure AI Search, MCP integration for Azure Boards, and CI/CD governance across 20+ microservices with JFrog promotion gates.
SPIDEY POWERS
🤖Agentic AI & LLMs
☁️Azure AI Platform
🏗️Cloud, IaC & DevOps
🐳Containers & Orchestration
🔎Vector & Search
🛡️Security & Observability
⚡Backend & APIs
🏅Quality Seals
READY TO GO AGENT-FIRST?
Email me, call me, or just say hi from anywhere in the multiverse. Your next production GenAI platform starts right here.