Priyabrat Dalbehera
Open to opportunities

GenAI Engineer LLM Pipelines & Agentic AI

I turn LLM research into production systems — RAG pipelines, agentic workflows, and AI-powered analytics built with LangChain, LangGraph, and Groq. 3+ years of real deployments, not just demos.

View My Work Talk to My AI →
Currently building
🔨  Building a self-correcting SQL agent with LangGraph tool-calling 📚  Learning AWS Bedrock & Claude API for enterprise RAG deployments 🧪  Experimenting with Agentic RAG — agents that rewrite their own queries
0 Projects Built
0 Years Experience
0 LLMs Integrated
📍 Bengaluru, India
GitHub LinkedIn
Priyabrat
Priyabrat Dalbehera GenAI Engineer · LLM & Agentic AI

GenAI Engineer with 3+ years building production-grade AI systems — from RAG pipelines and multi-agent orchestration to LLM deployments and revenue forecasting at scale.

🧠
Generative AI & LLMs
RAG systems, prompt engineering, agentic workflows with LangChain, LangGraph & CrewAI. Groq, OpenAI & Anthropic APIs.
🤖
Agentic Systems
LangGraph & CrewAI multi-agent workflows, SQL agents, tool calling, memory modules & self-correcting LLM loops.
📊
Data & AI Systems
Python, SQL, Pandas, Scikit-learn. Time series forecasting, ETL pipelines & KPI analytics for airline revenue systems.
☁️
Cloud & MLOps
AWS, Azure AI, AWS Bedrock, Docker. LLMOps — evaluation (RAGAS, LLM-as-Judge) & production deployment.
Capabilities

Everything I work with — in production, not just tutorials.

⛓️ LangChain 🕸️ LangGraph 🔍 RAG Pipelines 💾 ChromaDB 💾 FAISS 🧠 OpenAI API ⚡ Groq API ☁️ Azure OpenAI 🐍 Python 🦙 Ollama 📚 LlamaIndex 🔎 pgvector
🤖 Multi-Agent Workflows 🎯 ReAct Agents 🔧 Tool Calling 📊 RAGAS Evaluation ✍️ Prompt Engineering 🗄️ SQL Agents 👥 Human-in-the-Loop 🧠 LLM-as-Judge 🔁 Self-Correcting Loops 💬 Conversation Memory 🌐 MCP Protocol
⚡ FastAPI 🌶️ Flask 🐳 Docker 🦆 DuckDB 🔑 BM25 Retrieval 🎯 Semantic Search 📡 SSE Streaming 🧑‍🤝‍🧑 CrewAI 🔀 Hybrid Search 🗂️ SQL · DuckDB 👁️ GPT-4o Vision 📊 RAGAS
Technical Skills

My Stack

Specialized in GenAI engineering — from prompt design and RAG pipelines to production multi-agent systems.

🐍Python
⛓️LangChain
🕸️LangGraph
🔍RAG Systems
✍️Prompt Engineering
🧑‍🤝‍🧑CrewAI
💾FAISS / ChromaDB
⚡FastAPI
🐳Docker
☁️Azure OpenAI
🦆DuckDB
🗄️SQL
🤖 GenAI & LLMs
Function Calling Tool Use Agentic AI Embeddings Azure OpenAI Context Windows LlamaIndex Ollama AutoGen Anthropic API
🤖 Agentic Systems
LangGraph CrewAI Multi-agent Workflows Tool Calling SQL Agents Memory Modules Self-correcting Loops
🔎 Vector Search & Retrieval
ChromaDB FAISS DuckDB Pinecone pgvector BM25 Semantic Search Hybrid Retrieval
📊 Evaluation & MLOps
RAGAS LLM-as-Judge Hallucination Control Docker FastAPI Git
Projects

What I've Built

Production-grade AI systems — shipped, not just side projects.

01
🔗
✦ Featured

Advanced RAG Chatbot

A complete Retrieval-Augmented Generation system built from scratch, demonstrating end-to-end AI engineering — from raw document ingestion to evaluated, context-aware responses. Designed to highlight real-world architectural decisions, not just proto…

RAGCross-encoder re-rankingSentence-transformersChromaDBVector search
GitHub ↗ Case Study → Ask AI ↗
02
📈
Agentic AI

Finance AI Agent

Bull, Bear, and Risk Auditor agents debate stocks adversarially — not in parallel. A CIO judge delivers the verdict. Human approval required before the PDF report is generated.

LangGraphMulti-agentGroq·Llama 3.3OpenAI GPT-4oyfinanceTavily
GitHub ↗ Case Study → Ask AI ↗
03
🔍
LangChain

SalesQuery AI — Natural Language to SQL

An NL2SQL system that lets non-technical users query a sales database in plain English and get accurate SQL back — no schema knowledge required.

PythonLlama 3.2HuggingFace HubGradioSQLite
GitHub ↗ Demo ↗ Case Study → Ask AI ↗
04
📊
Agentic AI

Text2Insight — Natural Language Analytics Platform

A full-stack AI chatbot that converts plain English questions into SQL, executes them against a live sales database, and streams back data tables, charts, and LLM-generated summaries in real time. Features JWT authentication, multi-turn conversation …

HTML5CSS3Vanilla JSSSE StreamingFastAPI
GitHub ↗ Case Study → Ask AI ↗
05
🤖
LangChain

DataPilot-MCP

A conversational data analytics agent built on the Model Context Protocol (MCP). Ask questions in plain English — DataPilot queries your SQL database or CSV files, auto-generates charts, and returns narrative summaries with key insights.

MCPDUCKDB
GitHub ↗ Case Study → Ask AI ↗
06
👁️
LangChain

VisionRAG

A multimodal document intelligence system that processes scanned PDFs, invoices, and screenshots using GPT-4o Vision — extract structured data from visual documents and query them in natural language with source citations.

GPT-4o VisionpgvectorPostgreSQLFastAPIStreamlit
GitHub ↗ Case Study → Ask AI ↗
07
🔭
Agentic AI

LLMWatch

A plug-and-play LLM observability and evaluation platform for production agents. Drop-in callback handler wraps any LangChain or LangGraph agent in 2 lines — instantly tracks cost per query, latency percentiles (P50/P95/P99), token usage (input vs ou…

LangChainLangGraphRAGASLangSmithSQLite
GitHub ↗ Case Study → Ask AI ↗
Live Products

My Product

Real-world applications I've designed and shipped — click to try them live.

Loading…

Live Features

The AI Lab

Every tool below is live and running on this server — built to demonstrate production-grade AI engineering, not just theory.

🤖 Live

BM25 RAG Chatbot

Ask anything about my work. Keyword retrieval over portfolio knowledge base — sources + relevance scores shown on every response.

BM25RAGCitationsGuardrails
Open Chat →
🔗 Live

Multi-Agent Workflow

Watch 3 agents collaborate live — Researcher retrieves facts, Critic evaluates, Writer synthesises. Animated node graph with SSE streaming.

ResearcherCriticWriterSSE
Try It →
⚖️ Live

Model A/B Comparison

Send the same question to two models simultaneously. Compare latency, token count, and answer quality side-by-side in real time.

GroqOpenAIBenchmarking
Compare →
🧪 Live

Prompt Engineering Lab

Toggle between Zero-Shot, Few-Shot, Chain-of-Thought, and ReAct strategies — see exactly how prompt design affects LLM output quality.

Zero-ShotCoTReActFew-Shot
Open Lab →
📊 Live

LLM Stats Dashboard

Real-time usage metrics across all AI endpoints — requests, tokens, cost, and latency per model. Chart.js charts with 24h/7d/30d/1y views.

UsageLatencyCostCharts
View Stats →
🧠 Live

ReAct Research Agent

Multi-step agent running live: Plan → Research (BM25) → Synthesize. Built from scratch — no LangGraph dependency. Token-by-token streaming via SSE.

ReActPlanningSSENo LangGraph
Run Agent →
🏗 View Animated Architecture Diagram →
✦
Meet My AI

Say Hello to PRIYA

Personal Real-time Intelligent Yield Assistant — built to represent me 24/7 with intelligence that actually learns.

PRIYA is Live
11 GitHub repos synced
Always learning · 6 AI capabilities
🌐
Auto

Multilingual Mode

Replies in English, Hindi or any language the visitor uses — detected automatically, no setup needed.

Hindi Arabic Chinese +9 more
🧠
Always Learning

Self-Improving Memory

Stores frequent visitor questions and auto-improves answers over time using 👍👎 feedback loops.

BM25 RAG SQLite Feedback loops
🐙
Live Data

Live GitHub Sync

Auto-reads my latest GitHub repos and describes newest projects — without me updating anything.

GitHub API 1hr cache 17 repos
🎯
Smart

Visitor Intent Detection

Figures out if you're a recruiter, developer, or collaborator — and gives a tailored pitch automatically.

Recruiter Developer Collaborator
📅
Instant

One-Click Interview Booking

Shows a Google Calendar booking link the moment a recruiter says "Can we talk?" — instant meeting scheduling from chat.

Google Calendar 30-min call Any timezone
📲
Real-time

Telegram Live Alerts

Pings me in real time whenever someone starts a conversation — including what they asked, their city, and device.

Telegram Bot IP lookup Instant ping

Tap the ✦ button in the bottom-right corner or click above

Experience & Education

My Journey

From forecasting pipelines to production GenAI systems — 3+ years of real engineering, not side projects.

💼 Full-time Aug 2023 - Present
Anxilla Technologies Pvt. Ltd. (formerly Revemax) - Senior Software Engineer - GenAI & AI Systems

Anxilla Technologies Pvt. Ltd. (formerly Revemax) - Senior Software Engineer - GenAI & AI Systems

Architected and deployed production-grade GenAI systems using RAG, multi-agent workflows, and tool-calling LLM pipelines for airline revenue intelligence.

PythonLangChainLangGraphRAGLLMs (GPT/Azure OpenAI)FAISSChromaDBSQLDuckDBFastAPIDockerCloud Deployment
🧠
📊
💼 Full-time Jun 2022 - Aug 2023
Revemax (Rebranded to Anxilla Technologies Pvt. Ltd.) - Software Engineer - Airline Forecasting & Revenue Systems

Revemax (Rebranded to Anxilla Technologies Pvt. Ltd.) - Software Engineer - Airline Forecasting & Revenue Systems

Developed airline demand and revenue forecasting pipelines supporting commercial strategy and pricing decisions.

PythonSQLPandasSQL ServerScikit-learnForecasting ModelsData Pipelines
Open Source

GitHub Activity

Shipping consistently — every commit is a step forward.

@iampriyabrat14 ↗
⚡
– Events (90d)
📅
– Active Days
🔥
– Day Streak
🚀
– Last Push
Contribution Heatmap 90-day coding activity
Live Data
Mon
Wed
Fri
Loading contribution data…
Less More
Recent Commits Checking…
Live
Fetching activity…
View all on GitHub
Get In Touch

Let's Connect

Open to full-time GenAI Engineer roles, freelance projects, and research collaborations. Let's build something great.

Available for opportunities Open to Work
Email ipriyabrat689@gmail.com
→
LinkedIn priyabrat-dalbehera-p521
→
GitHub github.com/iampriyabrat14
→
📍Bengaluru, India
⚡Responds within 24 hours
🌐Remote & Hybrid ready
Send a Message
PRIYA
AI by Priyabrat
How would you like to connect?
Voice
Speak naturally
PRIYA talks back
Text
Type your message
Chat-style replies
GitHub syncing…
PRIYA
AI by Priyabrat
Live
Tap to speak with PRIYA
⌨ Switch to text mode
PRIYA
AI by Priyabrat
Switch to voice mode