Available · San Francisco, CA

Engineeringintelligenceinto production.

I'm Raju — an AI/ML Engineer with over 6 years across silicon, payments, and retail at scale. I build production LLM systems, agentic AI workflows, MCP servers, and RAG pipelines — and the evaluation and MLOps machinery that keeps them honest.

6+ years experience2× Cloud AI Certified (AWS GenAI Pro · Azure AI)MSCS · UMKC
Explore work
PythonFastMCPMCPLangGraphLangChainLlamaIndexMilvusFAISSPineconevLLMLoRAQLoRALlama 3GPT-5ClaudeAmazon BedrockCrewAIAutoGenLangSmithNeo4jSageMakerVertex AIBigQuerySnowflakeDatabricksPyTorchTensorFlowXGBoostKubeflowMLflowApache SparkPySparkRayAirflowKafkaFastAPITerraformKubernetesAWS GravitonEKSPythonFastMCPMCPLangGraphLangChainLlamaIndexMilvusFAISSPineconevLLMLoRAQLoRALlama 3GPT-5ClaudeAmazon BedrockCrewAIAutoGenLangSmithNeo4jSageMakerVertex AIBigQuerySnowflakeDatabricksPyTorchTensorFlowXGBoostKubeflowMLflowApache SparkPySparkRayAirflowKafkaFastAPITerraformKubernetesAWS GravitonEKS
Shipped at — past & present
ArmFIS GlobalWalmartArmFIS GlobalWalmartArmFIS GlobalWalmartArmFIS GlobalWalmartArmFIS GlobalWalmartArmFIS GlobalWalmart
01 · Positioning

Building AI systems
that are actually deployed.

Frontier models are easy to demo. Getting them past compliance, into production, and onto the critical path — that's the work I care about.

Philosophy

LLMs are a component.
MLOps is the product.

My work centers on the unglamorous layer between a model and a user — retrieval, evaluation, observability, safety rails, and deployment. I architect Model Context Protocol servers, LangGraph multi-agent workflows, and fine-tuned RAG pipelines that survive contact with auditors.

0%
Fewer hallucinations
0%
Faster onboarding
0%
Chargebacks cut
Currently

GenAI Engineer at Arm

Building the Arm MCP Server and the agentic retrieval stack behind it — helping developers move workloads from x86 to Arm using autonomous AI agents wired into Copilot, Claude Code, Kiro and Gemini CLIs.

Domains

Where I've shipped

  • Silicon & developer tooling
  • Payments & fraud
  • Retail supply chain
  • Search & recommendations
Signature stack

Agentic workflows

FastMCP · LangGraph · LlamaIndex · Claude · GPT-5 · Bedrock · Milvus · FAISS · vLLM · LangSmith

Cloud fluency

Multi-cloud by default

SageMaker · Bedrock · EKS · Vertex AI · BigQuery · Kubeflow · Databricks · Snowflake · Terraform

02 · Experience

Shipped in regulated
enterprise environments.

Three companies. Silicon, payments & fraud, retail supply chain at scale. Measurable outcomes at every stop.

01
Sep 2024 — Present
San Jose, CA
Arm logo
Arm

GenAI Engineer at Arm

Arm MCP Server & Cloud Migration Agents

Architected the Arm MCP Server and the agentic retrieval stack behind it — letting AI agents inside IDEs discover and chain cloud migration workflows that move production workloads from x86 to Arm across AWS Graviton, Google Axion and Azure Cobalt 100.

  • Architected the Arm MCP Server in Python + FastMCP so IDE agents discover and chain migration workflows — developer onboarding time down 60%.
  • Standardized MCP + JSON-RPC tool calling across GitHub Copilot, Claude Code, Kiro CLI and Gemini CLI, compressing migration analysis and refactoring for Graviton, Axion and Cobalt 100.
  • Built a containerized Milvus RAG knowledge base over Arm docs with LlamaIndex orchestration and LangSmith tracing — LLM code hallucinations down 85% at lower prompt-token cost.
  • Curated 12K instruction pairs from Arm docs and Neo4j graph triples and fine-tuned Llama 3 8B via LoRA (rank 16) — domain accuracy 52% → 70% on a held-out 200-query benchmark.
  • Benchmarked BGE-large, E5-large and text-embedding-3, then fine-tuned the selected bi-encoder with contrastive learning on hard negatives mined from failed retrievals — Recall@5 +15 points over baseline.
  • Added Cohere Rerank via Amazon Bedrock as a second stage over hybrid BM25 + dense retrieval, tuning chunk size and overlap for code-heavy docs — higher top-3 precision and fewer tokens per query for ~100ms added latency.
  • Orchestrated multi-agent workflows on GPT-5 and Claude via Bedrock with LangChain + LangGraph, served open-source models on vLLM for low-latency inference, and delegated agent roles through CrewAI and AutoGen.
  • Built Tesseract OCR pipelines to fold multi-format internal docs into the vector store, with Python jobs regenerating Milvus embeddings on every documentation update at zero user-facing downtime.
  • Embedded AWS Copilot CLI and Terraform into the retrieval layer behind guardrails, policy models and audit trails so autonomous agents emit only least-privilege IAM infrastructure.
  • LLVM-MCA accelerated assembly-level benchmarking 4×; sysreport and Arm Performix exposed as callable AI functions replaced 6 hours of manual telemetry with real-time profiling.
  • Shipped offline/online eval harnesses plus a CI pipeline running regression tests and prompt-version A/B comparisons before every release, with production observability on EKS and ECS tracking SLOs, agent health and telemetry.
PythonFastMCPMCPJSON-RPCLangChainLangGraphLlamaIndexLangSmithMilvusRAGCohere RerankGPT-5ClaudeAmazon BedrockLlama 3 8BLoRAQLoRAvLLMCrewAIAutoGenNeo4jTesseractTerraformAWS Copilot CLIEKSECSGravitonGoogle AxionAzure Cobalt 100LLVM-MCA
02
Jun 2022 — Jul 2023
India
FIS Global logo
FIS Global

Machine Learning Engineer at FIS Global

FraudSight for Worldpay

Real-time fraud detection and risk scoring across billions of payment transactions — end-to-end ML from large-scale feature engineering to sub-100ms inference, under PCI DSS, PSD2 SCA and 3-D Secure constraints.

  • Engineered the FraudSight real-time risk scoring engine (Adaptive Behavioral Analytics, scikit-learn, ensemble XGBoost) on billions of transactions — fraud chargebacks cut by up to 95%.
  • Designed feature engineering pipelines on Spark, PySpark and Databricks Lakehouse over 40B+ annual transactions, feeding Snowflake for identity profiling and model training.
  • Built RNN Deep Behavioral Networks (TensorFlow + Keras) on SageMaker to model cardholder transaction sequences, catching botnet attacks in real time while reducing false declines.
  • Deployed FastAPI + Featurespace ARIC inference microservices on EKS behind Java Spring Boot gateway services — sub-100ms scoring latency via optimized serialization and batch prediction paths.
  • Integrated predictive risk scores into the core payments gateway with Python and SQL, monitored via CloudWatch — securing billions in GMV at 99.99% uptime.
  • Automated retraining, hyperparameter tuning, experiment tracking and drift monitoring with SageMaker Pipelines, MLflow and Airflow for full regulatory auditability across model versions.
  • Defined typed API contracts and backend endpoints for the React + TypeScript analyst dashboard, surfacing live drift metrics, model version history and flagged transaction reviews.
  • Designed secure event-driven ML orchestration with Java, Spring Boot and Kafka — multi-threaded distributed services built for high reliability and horizontal scaling.
  • Aligned models to PCI DSS, PSD2 SCA and 3-D Secure, implementing PII controls and SHAP-based reason codes for declined transactions with full model lineage for audit.
PythonScikit-learnXGBoostTensorFlowKerasSageMakerSageMaker PipelinesApache SparkPySparkDatabricks LakehouseSnowflakeFastAPIJavaSpring BootKafkaEKSFeaturespace ARICMLflowAirflowCloudWatchSHAPPCI DSS
03
Aug 2019 — May 2022
India
Walmart logo
Walmart

Machine Learning Engineer at Walmart

Demand Forecasting & Catalog Intelligence

Owned demand forecasting and catalog enrichment for supply chain operations — end-to-end pipelines on GCP covering validation, feature generation, training, drift monitoring and automated retraining over terabyte-scale daily data.

  • Owned the end-to-end demand forecasting pipeline — validation, feature generation, training, deployment, drift monitoring and automated retraining — lifting forecast accuracy ~15% and cutting manual replenishment planning.
  • Contributed to a multi-modal catalog enrichment pipeline on GCP processing unstructured text and image data to close metadata gaps, supporting an estimated $20M+ annual GMV uplift.
  • Built time-series models (Prophet, ARIMA) on Vertex AI, orchestrating training, validation and experiment tracking through Kubeflow Pipelines and Airflow DAGs for automated retraining.
  • Trained distributed forecasting models across an estimated 2 TB of daily transaction and inventory data with PySpark, Dask and Ray — training time down ~40%.
  • Redesigned BigQuery partitioning and clustering across a multi-terabyte dataset — query costs down ~25% — and built a centralized feature store to standardize reuse across forecasting models.
  • Built Airflow, BigQuery and GCS workflows producing TFRecords for TPU training, integrating Multi-Task Learning frameworks to balance relevance against personalization.
  • Implemented label-aware BERT architectures to extract fine-grained product attributes from raw seller descriptions and customer reviews, cutting dependency on manual data across catalog pipelines.
  • Developed a computer vision pipeline (STN, CNNs, BiLSTM) to extract text and attributes from product images, resolving listings with missing or unstructured metadata.
  • Migrated keyword-based search to semantic search, contributing to indexing and retrieval optimization over millions of dense vectors on Milvus and FAISS.
  • Converted pandas workloads to PySpark for distributed processing, provisioned infrastructure with Terraform, and enforced code quality with pytest, Poetry and UV across parameterized, cache-aware KFP pipelines.
PythonPySparkDaskRayProphetARIMABERTCNNBiLSTMSTNVertex AIKubeflow PipelinesApache AirflowBigQueryGCSTFRecordsMilvusFAISSFeature StoreTerraformpytestPoetryUV
04
Feb 2019 — Aug 2019
India
Walmart logo
Walmart

Python Developer Intern at Walmart

Supplier & Inventory Data Pipelines

Data cleaning, ETL and validation for supplier and inventory datasets on the Hadoop stack, raising input data quality for downstream retail analytics workflows.

  • Built data cleaning and preprocessing scripts (pandas, NumPy) to standardize raw supplier and inventory datasets in Hive, improving input quality for downstream analytics.
  • Developed PySpark + HiveQL ETL pipelines on Hadoop for large-scale retail transaction data — processing time down ~30%.
  • Implemented automated validation checks in Python and SQL to flag supplier data inconsistencies, lifting data quality scores across Hive tables ~20%.
PythonpandasNumPyPySparkHiveQLHiveHadoopSQLETL
03 · Craft

The toolkit,
organized by layer.

156 / 156

Generative AI & Agents

42

Generative AI & LLMs

Frontier models + fine-tuning
  • GPT-5
  • Claude
  • Amazon Bedrock
  • Azure OpenAI
  • Llama 3 8B
  • Hugging Face
  • Transformer Architecture
  • LoRA
  • QLoRA
  • vLLM
  • Inference Optimization
  • Token Usage Optimization

Agents & RAG

Orchestration + retrieval
  • LangChain
  • LangGraph
  • LlamaIndex
  • FastMCP
  • MCP
  • JSON-RPC
  • RAG Architectures
  • CrewAI
  • AutoGen
  • Tool Calling
  • Cohere Rerank
  • Hybrid BM25 + Dense
  • Chain-of-Thought
  • Neo4j

Vector & Retrieval

Semantic search infrastructure
  • Milvus
  • FAISS
  • Pinecone
  • Azure AI Search
  • Semantic Embeddings
  • Vector Indexing
  • Contrastive Fine-Tuning

Evaluation & Observability

Tracing, benchmarks, monitoring
  • LangSmith
  • Braintrust
  • Model Evaluation
  • Offline / Online Evals
  • Prompt A/B Testing
  • MLflow
  • Prometheus
  • AWS CloudWatch
  • SLOs / SLIs

Machine Learning & Data

44

ML Frameworks

Classical + deep learning
  • PyTorch
  • TensorFlow
  • Keras
  • Scikit-learn
  • XGBoost
  • LightGBM
  • BERT
  • RNN
  • CNN
  • BiLSTM
  • STN
  • Prophet
  • ARIMA
  • Isolation Forest
  • SVM
  • Anomaly Detection
  • Multi-Task Learning

MLOps

From training to production traffic
  • MLflow
  • SageMaker Pipelines
  • Kubeflow Pipelines
  • Apache Airflow
  • Feature Store
  • Drift Monitoring
  • Hyperparameter Tuning
  • Databricks AI Gateway
  • Argo CD

Data Engineering

Pipelines at scale
  • Apache Spark
  • PySpark
  • Apache Airflow
  • Apache Kafka
  • pandas
  • NumPy
  • Dask
  • Ray
  • TFRecords
  • Parquet
  • HiveQL
  • Hadoop
  • Feature Engineering

Data Platforms

Warehouses + lakehouses
  • Snowflake
  • BigQuery
  • Databricks Lakehouse
  • Hive
  • Delta Lake

Cloud & Infrastructure

27

AWS

Core services used in production
  • SageMaker
  • Amazon Bedrock
  • EKS
  • ECS
  • Graviton
  • IAM
  • CloudWatch
  • AWS Copilot CLI
  • S3

Azure

Azure AI + compute stack
  • Azure OpenAI
  • Azure AI Search
  • Document Intelligence
  • AKS
  • Azure Cobalt 100

GCP

Google Cloud services
  • Vertex AI
  • BigQuery
  • GCS
  • Google Axion
  • Cloud Deployment Manager
  • TPU Training

Infra & CI/CD

Platform + delivery automation
  • Docker
  • Kubernetes
  • Terraform
  • GitHub Actions
  • Argo CD
  • CI/CD Pipelines
  • Prometheus

Engineering & Practices

43

AI-Assisted Development

Agents in the developer loop
  • Claude Code
  • GitHub Copilot
  • Cursor
  • Kiro CLI
  • Gemini CLI
  • AI Code Review
  • Security Screening

Languages

Day-to-day and systems
  • Python
  • Java
  • TypeScript
  • Go
  • SQL
  • Bash

Backend

APIs, protocols, async runtimes
  • FastAPI
  • Flask
  • asyncio
  • JSON-RPC
  • REST
  • Spring Boot
  • Apache Kafka
  • pytest
  • UV
  • pre-commit
  • Git

Safety & Compliance

Non-negotiables in regulated orgs
  • Prompt Injection Safeguards
  • PII Controls
  • Least-Privilege IAM
  • Regulatory Auditability
  • Data Governance
  • Secure Coding Practices
  • PCI DSS
  • PSD2 SCA
  • 3-D Secure
  • SHAP Reason Codes

OS & IDE

Working environment
  • Linux
  • Ubuntu
  • MacOS
  • Windows
  • Cursor
  • VS Code
  • Jupyter Notebook
  • JupyterLab
  • CLI tools
04 · Credentials

Certified on the frontier.

AIP-C01
AWS Certified Generative AI Developer badge

AWS Certified
Generative AI Developer — Professional

Amazon Web Services
AI-103
Microsoft Certified Azure AI Apps and Agents Developer badge

Microsoft Certified
Azure AI Apps & Agents Developer Associate

Microsoft
Education
Master of Science, Computer Science
University of Missouri — Kansas City
Aug 2023 — May 2025
GPA
3.82/4.0
05 · Let's collaborate

Have a hard problem
at the intersection of
AI and production?

Based in San Francisco and open to relocation. I'm open to AI/ML engineering roles in applied GenAI, agentic systems, and MLOps — especially where compliance, evaluation, and scale are non-negotiable.

Email
gkrswe@gmail.com
Phone
+1 (470) 242-1412
Network
linkedin.com/in/gkrswe
Code
github.com/gkr5413
© 2026 Raju Gottumukkala — AI/ML Engineer
Crafted with intention.