Open to senior AI/ML and GenAI engineering roles

I build AI systems that survive production.

Senior AI/ML Engineer with 7+ years across LLM applications, retrieval systems, retail forecasting, MLOps, distributed data pipelines, and cloud-native deployment across GCP and AWS.

Start a conversation Explore selected work

Google Cloud Professional ML Engineer credential earned Google Cloud Professional Cloud Architect credential earned AWS ML deployment experience

Based in Jammu, J&K

7+years in AI/ML engineering

2professional GCP credentials earned

5production domains: LLM, retail forecasting, vision, data, MLOps

Top 11%Kaggle competition finish

01 / Expertise

From model behavior to dependable systems.

I work across the full engineering path: retrieval and model orchestration, data pipelines, APIs, infrastructure, deployment, monitoring, and optimization.

LLM & retrieval systems

Grounded assistants, document ingestion, chunking, embeddings, vector search, citations, structured outputs, and evaluation-aware RAG workflows.

Cloud & MLOps

Repeatable training and inference systems with CI/CD, containers, autoscaling services, GPU workloads, observability, and reliability controls.

Retail forecasting & applied ML

Retail demand forecasting, time-series modelling, distributed feature pipelines, customer analytics, churn and delinquency models, and production inference.

02 / Selected work

Systems shaped by real constraints.

Selected work across production AI, data movement, retrieval, and deployment. Client-sensitive implementations are described without exposing private source or data.

Multi-cloud GenAIPublic reference implementation

AWS Bedrock RAG API reference implementation

Implemented a production-oriented FastAPI service for document retrieval and grounded generation on AWS, with S3 ingestion, Bedrock Knowledge Bases, least-privilege IAM, container delivery, structured logging, citations, tests, and deployment guidance. This is independent portfolio engineering and is not presented as client work.

View repository

FastAPIBedrockS3IAMDockerCloudWatch

Secure agent toolingPublic reference implementation

Secure multi-connector MCP gateway

Extended a Cloud Run-ready MCP gateway with OIDC authentication, role-based authorization, constrained BigQuery and PostgreSQL tools, governed REST connectors, PII masking, rate limits, and structured audit logging. Caller identity remains separate from the cloud workload identity.

View repository

MCPOIDCRBACBigQueryPostgreSQLPII masking

RAG evaluationPublic reference implementation

RAG Quality Workbench

Built a full-stack RAG workbench with swappable OpenAI, Claude, and Gemini providers; pgvector hybrid retrieval; reranking; citations; abstention; and latency, token, and cost traces. Its evaluation harness compares configurations on recall@k, MRR, keyword coverage, and abstention using the same pipeline as the live API.

View repository

FastAPINext.jspgvectorHybrid retrievalRerankingEvaluation

Agent reliabilityPublic reference implementation

Reliable LangGraph support agent

Built a router/supervisor graph with knowledge and order agents, persistent state, structured outputs, human approval for risky actions, low-confidence escalation, retries, timeouts, circuit breaking, full-round-trip tracing, and scenario-based evaluation through the compiled graph.

View repository

LangGraphHuman-in-the-loopFastAPIResilienceEvaluation

Semantic intelligencePrivate organization work

Policy intelligence and semantic search platform

Engineered a production platform for exploring policy records through semantic retrieval, combining a Dash and Flask application with Vertex AI embeddings, ScaNN search, and Cloud SQL. Delivered the supporting API and scheduled data workflows on Cloud Run with private database connectivity, Secret Manager, Artifact Registry, and workload identity federation.

Source private by design · Architecture discussable in interviews

Vertex AIScaNNCloud SQLCloud RunDash

Production GenAIIndependent engineering

Grounded RAG and intelligent assistant systems

Designed document-aware assistants spanning ingestion, chunking, embedding generation, retrieval, citations, response structure, and cloud deployment. Focused on grounded answers and operational reliability rather than demo-only prompting.

PythonGPT-4RAGGCP

Retrieval infrastructurePublic repository

Vector embeddings ingestion pipeline

Built a Cloud Run service that reads source records from Cloud SQL, generates embeddings with Vertex AI, and writes the enriched records back to a dedicated database table.

View repository

Cloud SQLVertex AICloud Run

Data engineeringPublic repository

GCS to BigQuery delivery workflow

Automated delivery of Python notebooks and ingestion assets across development, pre-production, and production GCP environments using service accounts and a GitHub pipeline.

View repository

GCSBigQueryCI/CD

Interview engineeringGrowing in public

Python interview solutions

A growing set of interview problems documented with intuition, dry runs, complexity analysis, clean Python, and automated tests—not answer dumps.

View repository

PythonDSATesting

03 / Experience

Seven years across the AI delivery lifecycle.

2026 — Present

Independent AI/ML Engineer

Upwork · Remote

Building GPT-powered automation, RAG platforms, assistants, GCP pipelines, GPU workloads, analytics integrations, and production Python services.

2023 — 2026

Machine Learning Engineer (ML/MLOps)

Open Insights · US, remote

Contributed to the Open ML 1.5 forecasting platform; built PySpark feature pipelines and TensorFlow training and inference workflows across BigQuery, Dataproc, and GCP infrastructure.

2022 — 2023

MLOps Engineer

Capgemini · Bengaluru

Deployed ML systems on GCP, moved architectures toward serverless delivery, and introduced carbon-emission tracking for ML workflows.

2019 — 2022

Data Scientist / ML Engineer

Hughes Systique · Gurugram

Led GCP and AWS model deployment, developed vision, NLP, and tabular models, optimized TensorFlow for mobile edge use, and maintained internal CI/CD tooling.

2019

ML Intern

Integration Wizards · Bengaluru

Trained TensorFlow and PyTorch models, prototyped pose estimation, and helped deploy a chemical-fluid leak detection system.

04 / Credentials

Certified depth, practical range.

B.Tech in Computer Science & Engineering from Shoolini University, with an OGPA of 8.14.

Google CloudProfessional Machine Learning Engineer credential earned

Google CloudProfessional Cloud Architect credential earned

NVIDIAJetson AI Certification

HackerRankPython Certification

05 / Recommendation

Direct manager feedback from production ML work.

A LinkedIn recommendation from a former manager who worked with me directly at Open Insights.

Bill Karr · Chief Scientist, Open Insights

Managed Raghu directly on the Open ML 1.5 platform

LinkedIn recommendation

Raghu was a core contributor to Open ML 1.5, our hierarchical probabilistic forecasting platform serving national retail clients at scale. Over the course of our work together he became genuinely well-rounded, comfortable moving between training and deploying ML models, writing and optimizing PySpark jobs, and managing cloud infrastructure on GCP, including BigQuery and Dataproc.

Production ML Model training, deployment, and practical delivery

Cloud engineering GCP, BigQuery, Dataproc, and clean repo discipline

Team trust Collaborative, dependable, and able to take ownership

06 / Contact

Have a production AI problem worth solving?

I’m interested in senior AI/ML, GenAI, MLOps, and cloud engineering opportunities where reliability matters as much as model capability.

raghusharma0404@gmail.com LinkedIn ↗ GitHub ↗ Upwork ↗