Building scalable AI systems with PyTorch, LLMs, RAG pipelines, and cloud-native infrastructure
AI Engineer with 3+ years of production experience developing and deploying machine learning systems, generative AI applications, and MLOps infrastructure. Currently pursuing M.S. in Artificial Intelligence at SUNY Buffalo while building enterprise-scale AI solutions at Cisco Systems.
Core Expertise:
- π§ Generative AI & LLMs: RAG pipelines, prompt engineering, fine-tuning (LoRA/PEFT), multimodal AI
- ποΈ Computer Vision: PyTorch model development, 3D reconstruction, deepfake detection, visual event analysis
- π§ MLOps & Production: Docker, Kubernetes, CI/CD, model serving (vLLM, Triton), monitoring (MLflow, Prometheus, Grafana)
- βοΈ Cloud & Infrastructure: AWS (SageMaker, Bedrock, EC2, S3), Terraform, distributed training, scalable microservices
Impact Delivered:
- Improved model accuracy by 18-30% across network automation and knowledge retrieval systems
- Reduced deployment cycles by 40% through automated ML/LLM pipelines
- Achieved 98% deepfake detection accuracy with Ξ²-VAE anomaly detection
- Built production CV pipeline with 90% violation detection across 300K+ parcels
graph TB
subgraph Data["π Data Layer"]
A1[PostgreSQL]
A2[MongoDB]
A3[Redis Cache]
A4[Vector DBs]
end
subgraph ML["π€ ML/AI Services"]
B1[PyTorch Models]
B2[LLM Inference]
B3[RAG Pipeline]
B4[Computer Vision]
end
subgraph Serving["β‘ Model Serving"]
C1[vLLM]
C2[Triton Server]
C3[FastAPI]
end
subgraph MLOps["π§ MLOps"]
D1[MLflow Tracking]
D2[Airflow Orchestration]
D3[Docker/K8s]
D4[CI/CD Pipeline]
end
subgraph Monitoring["π Observability"]
E1[Prometheus]
E2[Grafana]
E3[Model Drift]
E4[LangSmith/RAGAS]
end
subgraph Cloud["βοΈ Cloud Infrastructure"]
F1[AWS SageMaker]
F2[AWS Bedrock]
F3[S3/EC2]
F4[Terraform IaC]
end
Data --> ML
ML --> Serving
Serving --> MLOps
MLOps --> Monitoring
Cloud --> Data
Cloud --> ML
Cloud --> Serving
style Data fill:#e1f5ff
style ML fill:#fff3e0
style Serving fill:#f3e5f5
style MLOps fill:#e8f5e9
style Monitoring fill:#fff9c4
style Cloud fill:#fce4ec
Jan 2026 β Present
Generative AI & LLM Systems:
- Architected production RAG pipelines using LangChain, LangGraph, Hugging Face Transformers with semantic search and vector databases, achieving 30% improvement in contextual response accuracy
- Developed MCP-based integrations connecting LLMs with enterprise tools for tool calling and workflow automation
- Implemented LLM evaluation framework using RAGAS, LangSmith, TruLens for quality monitoring and retrieval performance
Computer Vision & Multimodal AI:
- Built multimodal AI applications using PyTorch with fine-tuning techniques (LoRA, PEFT) for image analysis, increasing inference accuracy by 20%
- Optimized model serving with vLLM and Triton Inference Server for production workloads
MLOps & Production Infrastructure:
- Engineered end-to-end ML/LLM pipelines using MLflow, Airflow, Docker, Kubernetes, GitHub Actions, reducing release cycles by 40%
- Developed scalable AI microservices with FastAPI, Redis, AWS (SageMaker, Bedrock) using Terraform IaC
- Implemented comprehensive observability with Prometheus, Grafana for latency, drift, and service reliability monitoring
Machine Learning & Optimization:
- Developed predictive models using PyTorch, TensorFlow, SQL, improving accuracy by 18% for network automation
- Optimized large-scale data processing with Apache Spark, PostgreSQL, reducing training time by 25%
AI Security & Responsible AI:
- Implemented guardrails, prompt-injection mitigation, access controls, and data protection across production services
Technologies: Python, PyTorch, TensorFlow, LangChain, LangGraph, Hugging Face, RAG, LoRA, PEFT, vLLM, Triton, MLflow, Airflow, Docker, Kubernetes, FastAPI, Redis, AWS (SageMaker, Bedrock, S3, EC2), Terraform, Prometheus, Grafana, RAGAS, LangSmith, Apache Spark, PostgreSQL, MCP
Jul 2022 β Jun 2024
Machine Learning Development:
- Developed predictive ML models using Python, TensorFlow, PyTorch, Scikit-learn, SQL, improving forecasting accuracy by 17%
- Optimized production models through feature engineering, hyperparameter tuning, and inference optimization, increasing accuracy by 15%
Data Engineering & Pipelines:
- Built ETL and feature engineering pipelines using Apache Spark, PySpark, PostgreSQL, reducing data preparation effort by 45%
MLOps & Deployment:
- Deployed ML services using FastAPI, Docker, AWS with CI/CD, reducing deployment time by 30%
- Implemented MLflow, Airflow workflows for experiment tracking, model versioning, and automated deployment
- Built scalable ML microservices using FastAPI, Docker, Kubernetes, Redis for concurrent inference workloads
Monitoring & Reliability:
- Implemented model monitoring with MLflow, Prometheus, Grafana tracking performance, latency, and drift, maintaining 95%+ stability
- Applied ML testing, data validation, and security practices using Pytest, Git, Docker, AWS
Technologies: Python, TensorFlow, PyTorch, Scikit-learn, SQL, Apache Spark, PySpark, PostgreSQL, FastAPI, Docker, Kubernetes, Redis, AWS, MLflow, Airflow, Prometheus, Grafana, Pytest, Git, CI/CD
State University of New York at Buffalo | Jan 2025 β May 2026
Mahindra University, Telangana, India | Aug 2020 β May 2024
mindmap
root((Gayathri Adulla<br/>AI Engineer))
Generative AI
LLMs & Transformers
Hugging Face
OpenAI GPT
Fine-tuning LoRA PEFT
RAG Systems
LangChain
LangGraph
Vector DBs
Semantic Search
Prompt Engineering
Context Engineering
Tool Calling
Structured Outputs
Computer Vision
PyTorch
TensorFlow
Multimodal AI
Object Detection
3D Reconstruction
Image Analysis
MLOps & DevOps
Pipelines
MLflow
Airflow
CI CD
Deployment
Docker
Kubernetes
FastAPI
Monitoring
Prometheus
Grafana
Model Drift
Serving
vLLM
Triton
Cloud & Infrastructure
AWS
SageMaker
Bedrock
EC2 S3
IaC
Terraform
Distributed
Spark
Ray
DeepSpeed
Data Engineering
Databases
PostgreSQL
MongoDB
Redis
Processing
Apache Spark
PySpark
ETL ELT
Vector Search
ChromaDB
FAISS
Pinecone
