Available for opportunities · Cincinnati, OH

ANVITA PANJUGULA

AI Scientist & GenAI Architect · RAG & LLM Systems in production at P&G

View My Work Let's Connect →
scroll to explore
3+ Years AI/ML
8+ Certifications
95% RAG Retrieval Acc.

Building the next layer of intelligence

I'm an AI Scientist — RAG & LLM Systems at Procter & Gamble's Digital Accelerator, where I architect production-grade agentic systems handling real enterprise workloads.

Currently completing my M.Eng. in Computer Science at the University of Cincinnati (April 2026), I specialize in multi-agent orchestration, hybrid retrieval pipelines, and making LLMs actually reliable in enterprise environments — with full governance, observability, and responsible AI compliance.

I shipped StatVisor — a multi-agent AI platform with BM25 sparse + dense hybrid retrieval, 95%+ accuracy, and 40% reduced API downtime. And I built Mockview, an AI mock interview platform, because I couldn't stop thinking about the problem.

Download Resume
India
2020
Cincinnati
2024
P&G
Now
Aug 2024 – Present
AI Scientist — RAG & LLM Systems
Procter & Gamble · Digital Accelerator
Aug 2024 – Apr 2026
M.Eng. Computer Science
University of Cincinnati
May 2023 – Jul 2024
Software Engineer — AI Automation
Analytics Quad4 · Bengaluru
Jun 2022 – Aug 2022
Software Engineer Intern — MLOps
Wissen Technology · Hyderabad
Aug 2020 – Jun 2024
B.E. Computer Science
Mahindra University · India

The stack I think in

From neural architecture to production APIs — the technologies I reach for when something needs to actually work.

Agentic AI / GenAI
LLM Engineering
LangGraph LangChain LlamaIndex RAG Pipelines Semantic Kernel MCP Guardrails AI Prompt Engineering Fine-Tuning / LoRA
Models & APIs
Foundation Models
Azure OpenAI GPT-4 Gemini Claude API Hugging Face Stable Diffusion ControlNet Transformers GraphQL
ML Frameworks
Model Development
PyTorch TensorFlow Scikit-learn XGBoost SciPy / NumPy CNN Reinforcement Learning CUDA
Backend
APIs & Services
FastAPI Python Asyncio Flask Node.js Spring Boot Kafka REST / JSON-RPC
Data & Vectors
Retrieval & Storage
FAISS Pinecone Weaviate pgvector Neo4j BigQuery Redshift Snowflake MongoDB PostgreSQL
MLOps & Observability
Production AI
Grafana Chainlit MLflow LangSmith Weights & Biases Airflow Docker Kubernetes Terraform
Frontend
Interfaces
React TypeScript Angular Vite Three.js Tailwind CSS Streamlit
Languages
Polyglot
Python R Java TypeScript C / C++ GoLang Rust MATLAB
Cloud & Infra
Infrastructure
AWS (SageMaker, EC2, EKS) Azure GCP Vertex AI Databricks CI/CD GitHub Actions Redis Spark

Things I actually built

Production systems and side projects — shipped, not just started.

Production 01

StatVisor

Multi-agent AI platform at P&G's Digital Accelerator. BM25 sparse + dense hybrid retrieval with OCR-based document ingestion; 95%+ retrieval accuracy across large enterprise datasets. Full LLM governance with audit logging, cost controls, schema validation, and real-time Grafana + Chainlit observability dashboards — reduced API downtime by 40%.

LangGraph LangChain LlamaIndex Azure OpenAI GPT-4 FAISS FastAPI MCP Neo4j Grafana Chainlit Semantic Kernel
Live 02

Mockview

Full-stack AI mock interview platform. Gemini AI adaptive question generation across 8+ categories with automated scoring on 5 dimensions. JWT auth with FastAPI/Supabase backend; trimmed API response times by 210ms and DB query times by 75%. Deployed via CI/CD on Render and Vercel with <2s load times.

React / Vite FastAPI Gemini AI Supabase PostgreSQL JWT Vercel Render
Research 03

NeuroCache

LLM memory optimization research project. Designed a memory management layer with context compression, FAISS-backed semantic prioritization, and fine-tuning strategies to reduce token usage while preserving reasoning quality. Async benchmarking engine across 10+ model configs on latency, memory, and accuracy.

Python PyTorch FAISS Asyncio LLMs Fine-Tuning
Research 04

Emotion Recognition

Real-time facial emotion classification using deep CNNs. TensorFlow/Keras model with OpenCV integration. Explored transfer learning and custom architecture design for affective computing.

TensorFlow Keras OpenCV CNN Python
Live 05

This Portfolio

Built from scratch — neural network canvas with Three.js, custom cursor, scroll-triggered animations, and an AI chatbot. Zero frameworks, pure HTML/CSS/JS.

Three.js Vanilla JS CSS Animations IntersectionObserver GitHub Pages

Where I've shipped

Aug 2024
Present
AI Scientist — RAG & LLM Systems (R&D)
Procter & Gamble — Digital Accelerator, Cincinnati, OH
  • Architected and shipped StatVisor, a production multi-agent AI system using LangGraph, LangChain, LlamaIndex, and Azure OpenAI GPT-4 — engineered function/tool calling, context grounding, BM25 sparse + dense hybrid retrieval, and OCR-based document ingestion with 95%+ retrieval accuracy across large enterprise datasets.
  • Built end-to-end LLM governance and observability infrastructure using FastAPI, Azure Monitor, and MCP — enforced audit logging, cost controls, schema validation, and responsible AI compliance; reduced API downtime by 40% via real-time Grafana and Chainlit monitoring dashboards.
  • Orchestrated multi-step agentic workflows using LangGraph, LlamaIndex, and Semantic Kernel with guardrails, multi-hop reasoning, and tool-calling agents; built knowledge graph retrieval pipelines using Neo4j and Stardog; aligned data governance with Collibra cataloging standards.
  • Engineered Snowflake Cortex LLM, vector search, and Cortex Analyst pipelines for AI-native retrieval; built async Python FastAPI backends across BigQuery and Redshift; developed statistical REST APIs using R Plumber, ggplot2, dplyr, and tidyverse — presented solutions directly to senior P&G leadership.
May 2023
Jul 2024
Software Engineer — AI Automation & Supply Chain
Analytics Quad4 — Bengaluru, India
  • Built and deployed GenAI agents and ML models in Python/R on GCP Vertex AI — applied hallucination mitigation strategies (output validation, grounding, confidence scoring) with feature engineering and model versioning across logistics, inventory, and supply chain planning workflows.
  • Designed high-throughput Python pipelines with concurrency and load-balancing, cutting processing overhead by 35% across millions of records; built multi-threaded Selenium scraper processing 11K+ sources in <3 seconds at 95%+ accuracy — adopted by 15+ engineers.
  • Orchestrated end-to-end ML workflows using Vertex AI, Airflow, MLflow, Docker, Kubernetes, Terraform, Spark, and Kafka; built Power BI dashboards with DAX and semantic layer modeling for supply chain KPI tracking.
Jun 2022
Aug 2022
Software Engineer Intern — MLOps
Wissen Technology — Hyderabad, India
  • Engineered end-to-end data workflows on Azure Cloud using ADF and Airflow — leveraged Azure compute and storage services with Docker-containerized services for scalable, production-grade data engineering and MLOps workloads.
  • Developed 3+ RESTful APIs and 10+ frontend components using Python, PostgreSQL, HTML, CSS, and JavaScript — streamlined data pipeline workflows serving 500+ daily transactions with 99%+ uptime.

Certified in what matters

☁️
Amazon Web Services
ML Engineer Associate (MLA-C01)
☁️
AWS
Certified AI Practitioner
🤖
NVIDIA
Deep Learning Professional
🧠
Anthropic
Prompt Engineering
🔥
Databricks
Generative AI Engineer Associate
❄️
Snowflake
Generative AI Professional
🏗️
IBM
RAG & Agentic AI Professional

Let's build
something real

Open to full-time AI/ML engineering roles & agentic system projects.

Ask Anvita anything ✦

Ask Anvita

● AI-powered · knows everything about me

Hi! I'm Anvita's AI assistant 👋 Ask me anything — about her projects, skills, experience, or if she's available for your role.