AI/ML Engineer

sairanga.mangina@gmail.com • +1 (864) 407-1871 • GitHub • LinkedIn

AI/ML Engineer with experience designing, fine-tuning, and deploying production-grade LLM, NLP, and Computer Vision systems serving 50K+ daily predictions across enterprise environments. Expert in PyTorch, TensorFlow, LangChain, Hugging Face, vector databases, and RAG pipelines, with a proven record of improving model accuracy by 19 points and reducing inference latency below 500ms. Skilled in MLOps (MLflow, Kubernetes, SageMaker, Docker), distributed data pipelines (Kafka, Spark, Airflow), and full-stack ML infrastructure on AWS, Azure, and GCP. Passionate about translating cutting-edge research into mission-critical AI products.

• Owned serverless backend on AWS Lambda, API Gateway, DynamoDB, and Step Functions; reduced cold-start fanout cost by ~22% and handled ~28K events/day.
• Designed GraphRAG platform on Neo4j with ColPali visual embeddings; indexed ~147K documents and achieved p95 retrieval of 480ms after adding a cross-encoder reranker.
• Fine-tuned Llama-3.1-8B and Mistral-7B with LoRA/QLoRA and MLflow on ~24K labeled specs; accuracy rose from 72% to 91% while 4-bit quantization cut inference cost ~45%.
• Built LiveKit + GPT-4o + Deepgram voice agent for Tier-1 support; deflection reached 63% and AHT dropped from 6 min to under 2 min.

• Built fraud-detection pipeline on SageMaker using XGBoost + Vision Transformer ensemble over ~240K geo-tagged submissions.
• Deployed PyTorch and scikit-learn models to EKS via MLflow + Seldon Core with Argo CD; cost per prediction dropped ~38% via async batch on the slow path.
• Implemented observability with OpenTelemetry, Datadog, and Evidently AI across 30+ features; MTTR improved from ~45 min to ~15 min.

• Built Django REST APIs to sync inventory and menus across Swiggy, Zomato, and Uber Eats for 500+ QSR clients.
• Rebuilt QA automation with RabbitMQ webhook integration, reducing flake rate to under 9% and eliminating manual rechecks.
• Delivered real-time analytics pipeline on Elasticsearch; P95 query latency improved from 3.0s to 1.2s via index redesign.
• Optimized Jenkins + Docker CI/CD with Python/Bash, cutting pipeline runtime from 18 min to 7 min.

• Delivered ASP.NET REST APIs for Razorpay and PayPal, handling ~25K monthly transactions with Polly circuit breakers.
• Migrated a C# monolith into five services (auth, payments, ledger, notifications, audit), reducing deployment time from 40 min to under 10 min.
• Authored Swagger/OpenAPI specs and runbooks; new hires became productive in week one.

• Integrated SAP ECC modules into ASP.NET REST services using Cloud SDK and IDoc protocols for plant operations sync.
• Built data validation that intercepted IDoc malformations before finance systems; saved hours of manual reconciliation.
• Created SQL KPI dashboards and shipped ABAP/BAPI enhancements for plant reporting.

• Built centralized React + Node admin panel for virtual classrooms, assignments, and digital attendance.
• Developed engagement dashboards and MongoDB + Express pipelines that surfaced grading discrepancies.

• AgentProbe — https://github.com/harsha-mangena/agent-probe
  AI-agent red-teaming framework with 106 attack vectors across 17 categories; SARIF-native CI/CD and compliance mappings (OWASP ASI 2026, MITRE ATLAS, EU AI Act, NIST AI 600-1).
• Agent Swarm — https://github.com/harsha-mangena/agent-swarm
  LangGraph + CrewAI multi-agent orchestration with shared vector memory and observability; deployed on FastAPI + Docker + AWS Lambda.
• Publications
  • Adaptive boosting with dynamic weight adjustment — https://arxiv.org/abs/2406.00524
  • Automated Data Preparation with Machine Learning for Enhanced Predictive Analytics — https://arxiv.org/abs/2407.00412

• Backend & Data: Python, Java, C#, JavaScript, SQL, Bash, FastAPI, Django, Flask, Spring Boot, ASP.NET, REST/gRPC, React.js, Node.js, PostgreSQL, MongoDB, Redis, MySQL, Elasticsearch, Neo4j, Pinecone
• LLM / Agents: RAG, GraphRAG, LangChain, LlamaIndex, LangGraph, CrewAI, LoRA, QLoRA, PEFT, Prompt Engineering, Vector Embeddings, OpenAI, Anthropic, Llama, Mistral, Hugging Face, ColPali, Transformers (BERT, GPT)
• ML / MLOps: PyTorch, TensorFlow, Keras, Scikit-Learn, XGBoost, LightGBM, MLflow, Seldon Core, Evidently AI, OpenCV, YOLO, NLP, Computer Vision, Generative AI
• Cloud & Infra: AWS (Lambda, SageMaker, EC2, S3, RDS, CloudWatch), GCP, Azure, Docker, Kubernetes (EKS), Terraform, Jenkins, Argo CD, Apache Kafka, RabbitMQ, Spark, Airflow
• Tools & Platforms: Git, GitHub, JUnit, Mockito, Pytest, ETL Pipelines
• Security & Compliance: OAuth2, RBAC, SOC 2, PCI-DSS

• M.S. in Computer Science, Clemson University — Dec 2024
  Graduate Student Teaching Assistant and Operations, Member of Association of Computing Machinery
• B.E. in Computer Science and Engineering, Aditya University — Aug 2021
  T-Hub intern, Representative of IEEE

AWS Cloud Practitioner • MS Azure Fundamentals • MS Database Administrator • AWS ML Fundamentals • GCP AI