KARTHIK ABHIRAM PIPPALLA

AI/ML · NLP · RAG · Backend Systems

Developing machine learning, NLP, RAG, and backend systems with Python, PyTorch, TensorFlow, FastAPI, and MongoDB — from model experimentation to production deployment.

Florida, USA
(754) 269-5640 · abhipippalla@gmail.com

LinkedIn Email View Projects

01

About

AI/ML engineer and data scientist with experience developing machine learning, NLP, RAG, and backend systems using Python, PyTorch, TensorFlow, Scikit-learn, FastAPI, and MongoDB. Worked across telecom analytics, healthcare NLP research, and end-to-end AI products involving model evaluation, retrieval, recommendation systems, and production deployment.

Experience spans model experimentation, evaluation, backend integration, and deployment of real-world AI applications — from 4G/5G anomaly detection and traffic forecasting at MasTec to hyperlocal community platforms, voice-memory apps, and medical chatbot research published at IEEE ACCAI 2024.

02

Skills

Programming & Tools

JavaScript Python Git Postman Swift

AI & ML

TensorFlow NumPy Scikit-learn Keras Pandas PyTorch OpenCV Transformers Agentic Workflows & MCP RNN CNN RAG Pipeline FAISS LoRA LangChain Whisper

Cloud & DevOps

Google Cloud Firebase Functions Firebase Cloud Tasks MongoDB Atlas AWS Docker

Data & Backend

MongoDB SQL REST APIs FastAPI Caching Strategies Kafka Redis Node.js Express.js

Security & Infrastructure

Rate Limiting Input Sanitization Session & State Management

Platforms

Telegram Bot API Hugging Face Google Cloud API Firebase n8n Xcode

03

Focus Areas

NLP & RAG Pipelines Retrieval & Recommendation Systems Telecom KPI Analytics Anomaly Detection & Forecasting Healthcare NLP Agentic Workflows & MCP ML Model Serving (FastAPI) Backend Integration & Deployment

04

Experience

  • MasTec India LLP

    Data Scientist

    January 2023 – August 2024

    Built 4G/5G anomaly-detection pipelines by cleaning KPI data with Pandas, engineering statistical features, and training Scikit-learn models to identify abnormal cell-site behavior, while using Git for version control and collaborative development relevant to active sprints.

    Developed traffic/capacity forecasting models from utilization, throughput, and time-based features to predict congestion-prone sites and support proactive planning.

    Implemented RF analysis across RSRP, RSRQ, SINR, latency, handover success rate, and PRB utilization, producing site-level performance indicators for optimization teams.

    Automated feature-engineering and ETL pipelines from SQL/network logs using Python/NumPy, including normalization, encoding, rolling averages, trend features, and validation.

    Created Isolation Forest and clustering workflows for unsupervised anomaly detection and built FastAPI inference services exposing trained models through REST endpoints.

05

Projects

Ibhood

Live

Hyperlocal community platform with rideshare, stays, HAP posts, and a natural-language Ask RAG pipeline. Creator & solo developer · 2026–Present.

Sourava — The Memory Assistant

Featured

iOS voice-memory app with on-device Whisper transcription, semantic indexing, and RAG Q&A.

RideEase

Multi-agent

AI-powered ride-sharing platform with conversational booking, geospatial RAG matching, and autonomous scheduling.

Event Dude

Winner

AI-powered event networking bot with multi-agent RAG matching, QR onboarding, and team formation. Miami Lamatic.ai Hackathon 2024.

Subway Voice Assistant

Recommendation AI

Voice-driven recommendation assistant combining speech recognition, preference modeling, and continuous improvement.

Drone Type Detection

Edge AI

Real-time edge detection for more than 200 drone categories, optimized and deployed on Raspberry Pi.

06

Publication

MEDOBOT · First Author · IEEE ACCAI 2024

Developed a medical chatbot for patient queries, medical question answering, appointment scheduling, and doctor recommendation. Fine-tuned and compared BERT and BioBERT using PyTorch and Hugging Face, with BioBERT outperforming BERT on the medical question-answering task.

Collaborated with doctors to collect, review, and validate medical question-answer data. Maintained separate training and benchmark datasets so evaluation examples were not exposed during model training. BioBERT achieved approximately 88% accuracy on the held-out evaluation dataset, with full system-level testing of response relevance and task completion.

View IEEE Publication

07

Education

  • Florida Atlantic University

    Master of Science, Computer Science (Artificial Intelligence)

    Boca Raton, FL, USA · 3.8 CGPA · May 2026

  • Sathyabama Institute of Science and Technology

    Bachelors of Science, Electronics and Communication (Elective - Artificial Intelligence)

    Chennai, TN, India · 8.8 CGPA · May 2024

08

Bookshelf

  • Deep Learning with Python by François Chollet
  • Make Your Own Neural Network by Tariq Rashid
  • Modern Robotics: Mechanics, Planning, and Control by Kevin M. Lynch and Frank C. Park

09

Contact

Florida, USA · (754) 269-5640 · abhipippalla@gmail.com