DK
Divakar Kumar
Available for AI/ML opportunities

Divakar Kumar

// AI/ML & Generative AI Engineer

AI/ML engineer with 4+ years specializing in LLM fine-tuning, multi-agent systems, and production-grade RAG pipelines. I build scalable AI solutions on Azure, Databricks, and modern agentic frameworks — turning complex data into actionable enterprise intelligence.

View ProjectsAsk my AI
4+
Years Experience
20+
AI Agents Deployed
70%
Analysis Time Reduced
5+
Advanced RAG Systems Deployed
~/divakar/agent.py
from langgraph import StateGraph
from agents import RAGSupervisor

graph = StateGraph(AgentState)
graph.add_node("retriever", retrieve)
graph.add_node("reasoner", reason)
graph.add_node("synthesizer", synth)

agent = RAGSupervisor(graph)
# → 70% faster decisions

// 01 — About

Engineering AI that ships, not slides.

Achievements

  • ★Awarded 'Best Team 2024' for AI-driven product delivery at Exponentia.ai
  • ★Mentored Sonata AI Innovation Team in intelligent agent development

Certifications

  • →DataCamp — Data Analysis with Python
  • →DeepLearning.AI — Neural Networks and Deep Learning
  • →Databricks — Generative AI Fundamentals
  • →Academy Europe — Introduction to AI

Education

  • Lovely Professional University
    B.Tech, Electrical & Electronics Engineering
    Aug 2017 — Jul 2020
  • St. Mary's College of Engineering & Technology
    Diploma, Electrical & Electronics Engineering
    Aug 2014 — Jul 2017

// 02 — Tech Stack

Tools I build with.

Production-tested frameworks and platforms across the modern AI stack.

Languages & APIs

PythonSQLNoSQLFastAPIREST APIPostman

GenAI / LLMs

Azure OpenAIGeminiAnthropicAmazon BedrockHuggingFaceMistralOllamaPrompt EngineeringFine-tuningvLLM

Agentic Frameworks

LangChainLangGraphLlamaIndexCrewAIMicrosoft Agent FrameworkGoogle ADKAgnoMCPA2A Protocol

ML / Deep Learning

Scikit-learnPyTorchTensorFlowKerasSpaCyNLTKFaster-Whisper

Vector & Databases

FAISSQdrantPineconepgvectorWeaviateMongoDBPostgreSQLMySQLRedis

Cloud & MLOps

AzureAWSGCPVertex AIDatabricksDockerKubernetesMLFlowCI/CDOpenTelemetry

Eval & Observability

LangfuseLangSmithRAGAsArize PhoenixPromptfoo

BI & Visualization

Qlik SensePower BIStreamlitGradioPandasNumPy

// 03 — Career

Experience timeline.

Aug 2025 — Apr 2026

Gen AI Developer

Sonata Software Limited

Hyderabad, India

Key contributions
  • ▸Designed enterprise-grade multi-agent AI systems with Microsoft Agent Framework, LangGraph & Copilot Studio.
  • ▸Built reference implementations comparing AutoGen vs MAF for architectural evaluation and migration.
  • ▸Implemented agent patterns: Agent-as-Tool, dynamic group chat, magnetic routing, swarm handoff, concurrent streaming.
  • ▸Established AI governance with Microsoft Purview — data classification, DLP, insider risk monitoring, audit trails.
Nov 2021 — Jun 2025

Engineer — BI

Exponentia.ai Pvt Ltd

Mumbai, India

Key contributions
  • ▸Developed an intelligent RAG-based system to generate context-aware responses by retrieving relevant information from both structured and unstructured data sources.
  • ▸Designed and implemented robust data pipelines for processing, chunking, embedding, and indexing large volumes of structured and unstructured data.
  • ▸Developed and deployed a multi-agent architecture (multi-agent supervisor AI) enabling collaborative agents to handle complex tasks like retrieval, reasoning, and decision-making over diverse data types.
  • ▸Delivered solutions in finance and NBFC domains, enabling better forecasting and decision-making.
  • ▸Designed and shipped 20+ interactive dashboards using Qlik Sense / Power BI; automated data workflows to drive insights for FinTech, NBFC, Agriculture, Chemicals and Agritech clients.
  • ▸Developed and deployed machine learning models to support data-driven use cases, integrating them into BI workflows for enhanced analytics.
  • ▸Led innovative POCs integrating Qlik AutoML and generative AI, enhancing forecasting accuracy and automating insight delivery.
Jul 2021 — Oct 2021

Data Analyst Intern

A for Analytics Pvt Ltd

Chennai, India

Key contributions
  • ▸Built Power BI dashboards for inventory & procurement analytics.
  • ▸Developed inventory prediction models, reducing overstock/understock by 20%.
  • ▸Automated ingestion pipelines, cutting manual reporting effort by 40%.

// 04 — Selected Work

Projects shipped.

Multi-Agent / FinTech

AI Loan Underwriting System

Production-ready underwriting with 5 DSPy ReAct agents for credit, income, affordability, compliance & decisioning — integrating OCR + RAG and real-time bureau data.

Impact
Cut underwriting from 3–5 days to <30s; ~40% better risk accuracy; ~80% lower ops cost with 100% auditability.
DSPyReActGPT-5.xBraintrustOCRRAGStreamlit
Agentic AI / Finance Ops

Multi-Agent Expense Management

Google ADK-orchestrated agents for expense validation, intelligent approval, and compliance auditing with configurable policy engines and real-time observability.

Impact
Approvals from 7 days → <2 min; ~94% cost reduction ($35 → $2); compliance 89% → 99%; 60% finance capacity gain.
Google ADKGemini 2.5 FlashGemini-3 ProAgentOpsPythonPolicy Engine
Enterprise Agentic AI

Production-Ready MAF Framework

Microsoft Agent Framework reference with swarm handoff, magnetic orchestration, dynamic speaker selection & agent-as-tool — across e-commerce, banking, travel & product domains with HITL.

Impact
~70% LLM cost reduction; latency 15–20s → 4–6s (sub-2s for critical flows); horizontally scalable & audit-ready.
Microsoft Agent FrameworkSwarmOrchestratorGPT-familyStreaming APIsPython
RAG / GenAI

Multilingual RAG Pipeline

Dual modular RAG pipelines unifying structured (SQL) and unstructured (PDF, Audio, Video) data for conversational analytics.

Impact
Reduced analysis time by 70% across departments.
PythonFastAPIFaster-WhisperAzure OpenAIGeminiLangChainMongoDB
Agentic AI

Advisr-X Multi-Agent System

Coordinated specialist agents for policy advice, comparison, pricing, and claims — simplifying insurance buying.

Impact
Improved decision accuracy and reduced research time via transparent comparisons.
PythonLangGraphLangSmithTavilyAzure OpenAIGemini
Multimodal

Invoice Extraction Chatbot

OCR + Gemini-powered chatbot extracting structured fields from multilingual, unstructured invoices.

Impact
Cut invoice processing time and entry errors by 60%.
PythonGeminiMistral OCRPDF2ImageDatabase
Enterprise AI

Multi-Agent System for Departmental Automation

Domain-specific supervised agents for PMO, Pre-Sales, HR & Marketing — automating reports, onboarding, proposals.

Impact
Eliminated silos and standardized decision-making across functions.
PythonLangGraphFastAPIAzure OpenAIPrompt Engineering

// 05 — Get in touch

Let's build something intelligent.

Open to AI/ML engineering roles, agentic systems consulting, and ambitious GenAI projects. Reach out — I usually reply within a day.

divakarkpandey@gmail.comLinkedInGitHub
+91 987-794-2954India