Forward Deployed AI Engineer

Jothiswaran Arumugam

Tallahassee, FL (Open to Relocate) 850-***-**** ********.***********@*****.*** linkedin.com/in/jothiswaran jo-portfolio-pi.vercel.app AI Engineer Forward Deployed AI Engineer

4+ years building production AI/ML systems for startup and enterprise clients across healthcare, financial services, and legal, currently the delivery lead at an AWS Advanced Tier Partner. I take projects from architecture scoping and prototyping through production deployment and live iteration, with 25+ discovery and delivery engagements to date. 13x AWS certified (Golden Jacket), 7x GCP certified. Technical Skills

Agentic AI & Tools: Claude Code, Strands Agents SDK, Bedrock AgentCore, LangChain, LangGraph, CrewAI, MCP AI/ML & LLMs: Claude, GPT-4, Llama, Mistral, PyTorch, Hugging Face Transformers, RAG and hybrid retrieval, QLoRA/LoRA fine-tuning, prompt and context engineering, LLM evaluation (RAGAS, LangSmith, DeepEval) Cloud & MLOps: AWS (SageMaker, Bedrock, Lambda, ECS Fargate, Step Functions, S3, DynamoDB, Redshift, OpenSearch, CloudWatch, CDK), GCP (Vertex AI, BigQuery, GKE, Cloud Run); OpenTelemetry, Terraform, Docker, Kubernetes, CI/CD (GitHub Actions) Engineering: Python, TypeScript, SQL, FastAPI, REST APIs Experience

Lead Data and AI Engineer Jun 2025 -- Present

RevStar Consulting (AWS Advanced Tier Partner) Tampa, FL

• Led technical discovery and requirements gathering across 25+ client engagements; translated business needs from non-technical stakeholders into production GenAI architectures including RAG pipelines, agentic workflows, and workflow automation solutions.

• Built a Text2SQL analytics agent using Claude and Bedrock AgentCore that converts natural language to Redshift SQL across a 50+ column schema with SQL injection prevention, PII guardrails, and multi-tenant isolation, now serving 300+ queries a week from a 14-analyst team.

• Reduced inference latency from 8.3s to 1.2s and cut costs by 93% by designing AI transaction categorization with Bedrock AgentCore and dual-namespace long-term memory; benchmarked Haiku 4.5 vs Sonnet for model optimization.

• Implemented QLoRA fine-tuning pipeline for Llama 3.1 8B on AWS SageMaker with 4-bit quantization and distributed training across 8x A100 GPUs, achieving 4-8x training speedup via BitsAndBytes NF4.

• Architected production RAG pipelines on AWS Bedrock and OpenSearch serving 10+ clients with hybrid retrieval; built proof-of-concepts to demonstrate feasibility and drove improvement in grounding accuracy across deployments.

• Cut delivery timelines by 70% by creating internal quickstart frameworks (Strands patterns, AWS CDK templates, CI/CD) and automated repetitive development workflows now used department-wide.

AI Engineer (Co-op) May 2024 -- May 2025

IM Digital Boca Raton, FL

• Built RAG pipelines with hybrid vector indexing on Qdrant achieving sub-300ms retrieval latency for document processing and data extraction across legal and product domains, serving 1K+ daily queries.

• Shipped multi-agent LLM APIs handling 1K+ daily queries on Cloud Run with FastAPI, RBAC, rate limiting, and Prometheus/Grafana observability; iterated based on end-user feedback and usage monitoring.

• Improved hallucination grounding scores by 62% in production by building CI/CD-driven LLM evaluation with LangSmith and RAGAS. GCP Cloud Engineer (AI/ML) Jul 2021 -- Aug 2023

Cognizant Technology Solutions Hyderabad, India

• Automated ML infrastructure provisioning using Terraform, GKE, and BigQuery, reducing setup time by 80% across 15+ projects.

• Deployed and managed ML model serving pipelines on Vertex AI with CI/CD automation, improving latency observability and SLA adherence by 60%.

• Optimized ETL workflows in Dataflow and BigQuery, cutting cloud compute costs by 30% and boosting daily pipeline throughput.

• Mentored 30+ engineers in GCP, ML infrastructure, and deployment best practices, accelerating internal cloud adoption by 4x. Projects

• AgentOps Lab (github.com/devjothish/agentops-lab): Framework-agnostic reliability toolkit that traces LangGraph agent trajectories via OpenTelemetry, scoring and visualizing each run to quantify the impact of agent changes.

• RAG Refusal Eval (github.com/devjothish/rag-refusal-eval): Golden-question harness over 1,591 indexed passages that refuses when its sources cannot support an answer, validating every claim against a verbatim retrieved quote instead of a similarity threshold.

• Agentic RAG with Qdrant and Tavily (github.com/devjothish/AgenticRAG): Multi-step agentic retrieval outperforming single-shot RAG on queries needing both indexed knowledge and live web context.

Certifications

AWS (13x, Golden Jacket): GenAI Developer Professional, Solutions Architect Professional, DevOps Engineer Professional, Advanced Networking Specialty, Machine Learning Specialty, Security Specialty, Developer Associate, Machine Learning Associate, Data Engineer Associate, Solutions Architect Associate, CloudOps Engineer Associate, AI Practitioner, Cloud Practitioner Google Cloud (7x): Professional Machine Learning Engineer, Professional Cloud Architect, Professional Data Engineer, Professional Cloud Developer, Professional DevOps Engineer, Professional Database Engineer, Associate Cloud Engineer Education

Florida State University Aug 2023 -- May 2025

Master of Science in Computer Science Tallahassee, FL Rajalakshmi Engineering College Aug 2017 -- May 2021 Bachelor of Engineering in Computer Science Chennai, India

Contact this candidate