MG Mounika Grandhi

AI Engineer · Hyderabad, India

Hello, I’m Mounika.
I build AI systems that ship.

Open to AI engineering roles

AI Engineer building agentic workflows, RAG systems, conversational systems, and production LLM backends across OpenAI, Anthropic, Gemini, Grok, AWS, and GCP.

Designing AI workflows that are grounded, observable, and reliable enough for real product teams.

View projects Resume
Agentic Workflows multi-agent systems, tool use, HITL flows, and long-term memory
Document Intelligence structured extraction, hybrid RAG, retrieval, grounding, and validation
Conversational Systems assistants, analytics workflows, and natural-language product interfaces
Production AI APIs, async pipelines, observability, evaluation, and deployment

01 · About

Useful, observable, grounded by design.

I’m an AI Engineer focused on building practical GenAI systems for real product workflows, especially document intelligence, retrieval, structured extraction, and backend orchestration.

I turn vague product ideas into AI systems with clear inputs, reliable outputs, measurable behavior, and APIs that teams can actually integrate. My work sits between LLM application engineering and backend software: context design, RAG, prompt workflows, async pipelines, validation layers, observability, and evaluation.

I care less about flashy demos and more about AI that survives production pressure: structured outputs, traceable behavior, human-review paths, failure handling, and enough instrumentation to debug and improve what ships.

02 · Selected Projects

Working architecture over polished demos.

01

Swaram AI

Open-source SDK for voice AI agents without a hosted orchestration platform.

A TypeScript SDK for building domain-agnostic voice support agents, coordinating VAD, STT, LLM reasoning, tool calls, TTS, transcripts, interruptions, and telephony transport.

  • TypeScript
  • Voice Agents
  • Groq
  • Ollama
  • Whisper.cpp
  • Kokoro TTS
  • Twilio
02

Content Automator

Agent-directed content pipeline with deterministic rendering and human review.

A Google ADK pipeline where an LLM Director chooses strategy and style, while deterministic Python services handle research, script generation, image sourcing, TTS, captions, FFmpeg assembly, Telegram review, and approval-gated publishing.

  • Python
  • Google ADK
  • Gemini
  • FFmpeg
  • Telegram Bot
  • YouTube API
03

Doc Extractor

Template-driven app that turns documents into confidence-scored structured data.

A full-stack document extraction app with custom templates, PDF/DOCX/image parsing, Groq-powered JSON extraction, confidence scoring, persisted results, batch jobs, and a Next.js interface.

  • FastAPI
  • Next.js
  • Groq
  • Vision LLM
  • SQLModel
  • Celery
04

CompIntel

Currently building
AI competitive-research desk for sourced market and competitor reports.

A multi-agent + RAG system that decomposes competitor research into analyst agents, retrieves evidence from documents and web search, verifies claims with citations, and routes final reports through a human approval gate.

  • LangGraph
  • Hybrid RAG
  • FastAPI
  • Chroma
  • HITL
  • Evals

03 · How I Build

Using AI with reliability, deployment, and observability.

How I Use AI

Use LLMs where they improve reasoning, retrieval, summarization, and workflow automation, with clear boundaries around tool use and human review.

Deployments

Ship AI services through backend APIs, Dockerized services, cloud integrations, provider routing, and reliable connections to product data and tools.

Monitoring, Evaluation & Observability

Track traces, latency, token usage, retrieval quality, grounding behavior, and evaluation signals so model behavior can be debugged and improved.

04 · Experience

Production GenAI across startups and enterprise modernization.

Jul 2025 · Present

AI Engineer · Inncircles

Agentic AI · RAG · Analytics
Tool integrations 4 LLM providers LangSmith tracing

Building production AI systems for construction workflows, with multi-agent orchestration, retrieval, model routing, external tools, and stakeholder-facing conversational experiences.

  • Built multi-agent AI workflows for construction operations, enabling automated analysis from workflow data.
  • Designed hybrid RAG with chunking, semantic retrieval, token-aware batching, and grounding guardrails.
  • Integrated MCP tools, human-in-the-loop flows, and long-term memory patterns across operational workflows.
  • Added multi-modal and multi-provider routing across OpenAI, Anthropic, Gemini with Pydantic outputs and LangSmith tracing.
Aug 2022 · Jun 2025

Digital Specialist Engineer · Infosys

Legacy modernization · RAG · Evaluation
Modernization AI 5+ codebases Evaluation pipelines

Shipped GenAI capabilities for enterprise modernization, helping teams understand, document, retrieve, and validate legacy system knowledge at scale.

  • Contributed to an AI migration platform for reducing manual effort in legacy modernization workflows.
  • Implemented RAG-based QA chains across 5+ legacy codebases for reverse engineering and knowledge extraction.
  • Built code summarization and BRD generation pipelines with LangChain, embeddings, and vector databases.
  • Designed evaluation pipelines using BLEU, CodeBLEU, and custom benchmarks to track output quality and hallucinations.

05 · Stack

The stack behind systems that ship.

Agentic AI

Core: LangChain · LangGraph · DeepAgents · Google ADK · MCP

RAG & Data

Core: LlamaIndex · Milvus · ChromaDB · Semantic Search
Used: MySQL · MongoDB

LLM Providers

Core: OpenAI · Anthropic Claude · Google Gemini · Grok
Used: Groq · HuggingFace

Backend & Cloud

Core: Python · FastAPI · REST APIs · Docker · AWS · GCP
Used: Java · RabbitMQ

Evaluation & LLMOps

Core: LangSmith · Prompt Engineering · Monitoring
Used: BLEU · CodeBLEU · CI/CD

06 · Contact

Let’s talk about AI systems that need to work in the real world.

Email GitHub LinkedIn Resume

Hyderabad, India · Open to AI engineering roles and production GenAI collaborations.