Samhita Kolluri

About Me

Boston, MA (Open to Relocation)

Building AI for Impact

AI Research Engineer bridging the gap between scalable Data Engineering and Agentic AI. Architected autonomous multi-agent systems and high-precision RAG pipelines for the Stellis Labs x Humanitrans AI collaboration. Previously spearheaded ETL optimization for Fortune 500 clients at Cognizant, delivering 99% efficiency gains by migrating legacy pipelines to vectorized SQL layers. I build fault-tolerant, production-grade systems that turn research into revenue.

Python GenAI & LLMs RAG Architecture Agentic Workflows PySpark Databricks Snowflake Docker Kubernetes Airflow

Contact

samhita.kolluri@gmail.com

Download Resume

Technical Skills

AI/ML & LLMs

Fine-tuning LLMs LangChain LangGraph Multi-Agents RAG Prompt Engineering Hugging Face BERT Transformers LLaMA

Deep Learning & Computer Vision

Computer Vision OpenCV CNN LSTM Image Processing MediaPipe TensorFlow PyTorch Reinforcement Learning

MLOps & Data Engineering

Docker Kubernetes Apache Airflow MLflow CI/CD GitHub Actions DVC PySpark Databricks DBT ETL Informatica PowerCenter

Cloud & Databases

GCP AWS Azure Snowflake Vector Databases ChromaDB MySQL PostgreSQL Cloud Firestore SQL

Languages & Frameworks

Python R SQL Java C++ FastAPI Streamlit React Spring Boot Microservices

Analytics & Visualization

Tableau PowerBI Data Analysis Data Visualization Statistical Analysis Time Series Analysis A/B Testing

Projects

PhysioPro

Generative AI Motion Correction Tool

Developed a physiotherapy assistance tool improving movement correction accuracy by 30% using Procrustes analysis and DTW. Integrated AI recommendations with Cortex Mistral 7B and Snowflake, boosting adherence by 25%.

Snowflake Cortex OpenCV MediaPipe Mistral 7B Streamlit

GitHub Medium

HomieHub

Production MLOps Architecture

Deployed production MLOps ecosystem on Google Cloud using Docker and Airflow. Created hybrid retrieval engine with custom LLM Agent to autonomously sanitize and analyze high-volume unstructured social data streams.

Google Cloud Run Apache Airflow Docker LangChain DVC

GitHub Demo

Contrastive Ideas Search Module

Advanced RAG Architecture

Built hybrid embedding framework leveraging mxbai-embed-large and fine-tuned LLaMA 3 model for semantic opposition detection. Implemented multi-vector indexing with ChromaDB for efficient retrieval.

LLaMA-3 ChromaDB Hugging Face Fine-tuning

GitHub HF Model Medium

Evil Geniuses

Adversarial Safety Research

Analyzed vulnerabilities in Multi-Agent Systems by mathematically modeling adversarial attacks. Investigated cascading failures and proposed System Role-Based Filters to mitigate risk.

Multi-Agent Systems Red Teaming LLaMA-3 CAMEL MetaGPT

Blog

SEMANTIC

Multi-class Text Classification

Developed multi-class classification system using IAB-labeled dataset. Fine-tuned transformer models for efficient context-aware classification in imbalanced data scenarios.

BERT PyTorch NLP Transformers

GitHub

SysTune

LLM-Based Hardware Optimization

Developed LLM-based autotuning system using GPT-4 to optimize hardware and software parameters, enhancing HPC resource utilization by 30%.

GPT-4 HPC Python OpenAI API

GitHub

View More on GitHub

Work Experience

AI Research Engineer

Humanitarians AI & Bear Brown Company

January 2025 - May 2025

Architected autonomous agentic system using LangChain and FastAPI with hybrid retrieval achieving 40% lower inference latency
Fine-tuned Llama-3 using QLoRA on custom datasets for domain-specific noisy data handling
Engineered persistent memory layers using LangSmith for hallucination-free production environments
Orchestrated scalable RAG pipeline with ChromaDB and cross-encoder re-ranking, increasing analysis accuracy by 35%

Graduate Teaching Assistant

Northeastern University

December 2024 - June 2025

Mentoring graduate students in Data Mining & GenAI applications
Designing interactive labs on data storytelling with Power BI and Python
Partnering with Prof. Mohammad Dehghani to modernize course materials for IE 5374

Senior AI Data Engineer

Cognizant Technology Solutions

August 2022 - July 2023

Led modernization of enterprise data infrastructure, achieving 99% acceleration in ETL reconciliation
Re-architected critical reconciliation workflows reducing processing time from hours to minutes
Designed automated governance frameworks achieving 100% compliance with audit standards

AI Data Engineer

Cognizant (Generative AI Research Lab)

March 2021 - August 2022

Deployed high-throughput PySpark pipelines on Databricks for LLM pre-training data preparation
Designed distributed transformation logic for "Gold Standard" training datasets
Refactored legacy ingestion scripts reducing latency by 20%

Computer Vision Research Intern

Bennett University

May 2020 - June 2020

Led development of Sign Language Recognition system achieving 98.56% validation accuracy
Implemented advanced image pre-processing pipelines using Kirsch compass masks
Fine-tuned facial expression models on Fer-2013 dataset achieving 86% accuracy

Education

MS in Data Analytics Engineering

Northeastern University

August 2023 - December 2025 | GPA: 3.8/4.0

Coursework: Machine Learning Operations (MLOps), Gen AI with LLM in Data Engineering, LLM-based Dialogue Agents, Natural Language Processing, Data Mining, Cloud Computing

BTech in Computer Science Engineering

VNR VJIET

2017 - 2021

Coursework: Artificial Intelligence & Neural Networks, Data Structures & Algorithms, Computer Graphics, IoT, Cognitive Science, Cyber Security

Research & Publications

An Artificial Intelligence and Internet of Things based Integrated Approach for COVID-19 Prevention

Patent Application Num: 202141054101, December 2022

Designed a hardware-integrated Computer Vision system for real-time monitoring. Optimized inference latency for low-compute devices (Raspberry Pi) using lightweight CNN architectures.

AI IoT MobileNetV2 Raspberry Pi

Post-Quantum Cryptography Framework

International Journal of Engineering Research & Technology (IJERT)

Developed a security framework analyzing lattice-based cryptographic resilience against quantum computing attacks. Modeled decryption vectors to harden digital infrastructure.

Cybersecurity Lattice Algorithm Quantum Resilience IJERT

Read Paper

AI-based Screening System for COVID-19

IEEE 7th International Conference for Convergence in Technology (I2CT), 2022

Developed a computer vision system for COVID-19 detection achieving 96% accuracy using deep learning models.

Paper

Brain-Computer Interface

Annual Technical Symposium, India, September 2021

Presented research on brain-computer interface technologies and applications.

Get In Touch

Let's Build Production-Grade Systems.

Whether you're looking to optimize your ETL workflows or architect a new Multi-Agent system, I'm always open to discussing complex engineering challenges.

Start a Conversation Schedule a Chat

Hello, I'm