Home

tharun cheruku - Data Science

[email protected]

Location: Milwaukee, Wisconsin, USA

Relocation: YES

Visa: OPT

Resume file: Tarun DS_1765461434819.docx Please check the file(s) for viruses. Files are checked manually and then made available for download.

Tarun Goud Cheruku Data Scientist Generative AI | AWS Bedrock | LLMs | NLP | AI Solutions E-Mail: [email protected] Phone: 414 797 8147 EDUCATION B Tech, Brilliant Group of Institutions, 2022 MS in Computer Information Science, Concordia University WISCONSIN PROFESSIONAL SUMMARY Data Scientist with 6+ years of experience specializing in Generative AI, LLMs, NLP, and end-to-end machine learning solutions. Expert in designing and deploying enterprise-grade GenAI applications using AWS Bedrock, SageMaker, and serverless AWS architectures. Strong proficiency in integrating Claude, Mistral, LLaMA, Titan, and GPT models for summarization, chatbots, document intelligence, and automation. Hands-on experience building Retrieval-Augmented Generation (RAG) systems using FAISS, Pinecone, Chroma, and Bedrock Embeddings. Skilled in LLM fine-tuning, parameter-efficient tuning (PEFT), prompt engineering, and hallucination-reduction techniques. Proficient in developing microservices and APIs using FastAPI, Flask, Lambda, and API Gateway for scalable model deployment. Demonstrated ability to optimize inference workloads through model routing, quantization, and autoscaling, reducing operational costs. Deep expertise in NLP pipelines, including tokenization, feature engineering, entity extraction, summarization, and semantic similarity scoring. Experienced in building automated data engineering pipelines using Glue, Step Functions, Spark, Redshift, and Delta Lake. Strong track record of improving model reliability using LangSmith, PromptLayer, CloudWatch, and SageMaker Model Monitor. Proven ability to collaborate with cross-functional teams to translate business goals into AI-driven, production-ready solutions. Adept at implementing secure deployments with IAM, VPC, KMS, encryption, and AWS governance best practices. Delivered multiple multi-agent GenAI frameworks for enterprise automation and scalable decision-support systems. Excellent communicator with experience presenting GenAI architectures, ROI insights, and adoption strategies to executive stakeholders. Passionate about responsible AI, model transparency, and delivering high-impact, domain-specific solutions across finance, healthcare, and public sectors. TECHNICAL SKILLS Programming: Python, SQL, Bash, JavaScript, TypeScript Frameworks: PyTorch, TensorFlow, Scikit learn, Hugging Face Transformers, LangChain, LangGraph GenAI & LLMs: AWS Bedrock (Claude, Mistral, LLaMA, Titan), GPT, RAG, Prompt Engineering Cloud: AWS (SageMaker, Glue, Lambda, S3, ECS, API Gateway, KMS), Azure, GCP Data Engineering: Glue, Redshift, Step Functions, Snowflake, Databricks, Spark Vector DBs: FAISS, Pinecone, Chroma, Weaviate APIs & Microservices: REST, FastAPI, Flask, API Gateway, Boto3 Monitoring: LangSmith, PromptLayer, W&B, CloudWatch, Model Monitor Security: IAM, KMS, VPC CI/CD: GitHub Actions, Jenkins, MLflow, Terraform, Docker, Kubernetes Visualization: Power BI, QuickSight, Grafana Soft Skills: Leadership, stakeholder collaboration, Agile, mentoring PROFESSIONAL EXPERIENCE Client: Citi, California Lead Data Scientist Generative AI (AWS Bedrock & SageMaker) Feb 2024 Present Designed and deployed large-scale GenAI systems supporting financial analysis, compliance workflows, and automated reporting across high volume enterprise environments. Architected an end-to-end RAG ecosystem using Bedrock models, FAISS indexes, DynamoDB metadata stores, and Lambda-based orchestration, improving document discovery speed by 60%. Led LLM fine-tuning initiatives using domain specific financial datasets, improving contextual accuracy for sentiment extraction, portfolio summarization, and risk scoring tasks. Built fully containerized chatbot microservices using Lambda, API Gateway, and ECR, servicing 50K+ queries weekly with sub second response times. Designed automated prompt evaluation pipelines using LangSmith, enabling continuous monitoring of hallucination rates, factual accuracy, and cost metrics. Drove cost optimization strategy across inference workloads using model routing, quantization, and autoscaling, resulting in a 25% reduction in spend. Partnered with enterprise security teams to implement IAM based RBAC, KMS encryption, PrivateLink, and governance controls for LLM deployment. Presented GenAI roadmap, ROI metrics, and architectural recommendations to senior leadership, influencing multi million dollar modernization decisions. Client: USPS, Texas Senior Data Scientist AWS GenAI Solutions Dec 2022 Feb 2024 Led the development of document intelligence solutions using Bedrock LLMs, reducing manual review effort by 40% across logistics and compliance teams. Built scalable SageMaker-hosted LLM inference endpoints supporting multi tenant workloads; implemented canary deployment strategies to minimize downtime. Designed multi agent automation frameworks (intent classification, summarization, verification agents) for optimizing internal communications processing. Implemented comprehensive prompt-guardrail systems combining regex filtering, embedding based toxicity checks, and AWS content moderation APIs. Developed automated training pipelines using SageMaker Pipelines with integrated CI/CD deployments via GitHub Actions. Built CloudWatch-driven LLM observability dashboards for latency, token usage, cost trends, and anomaly detection. Conducted internal training workshops on Bedrock and LangChain to enable adoption across federal teams. Client: Business Object Solutions, Virginia Data Scientist Generative AI Sep 2021 Nov 2022 Built production-grade GenAI assistants, summarizers, and retrieval systems using Bedrock + Hugging Face transformers. Developed prompt tuning frameworks with structured templates, dynamic routing logic, and context aware injection strategies. Engineered automated data ingestion and preprocessing pipelines involving chunking, embeddings generation, and metadata extraction using AWS Glue and Step Functions. Built semantic search and document QA systems using Pinecone vector DB, improving search precision for internal users. Designed A/B testing harnesses for evaluating model performance, style adherence, and hallucination rates across model families. Implemented model compression, quantization, and autoscaling techniques to reduce inference cost by 30%. Trained teams on best practices for RAG design, embeddings selection, and evaluation methodologies. Client: State of Kansas, Kansas AI/ML Developer July 2019 Aug 2021 Built scalable ML models for fraud detection, eligibility scoring, and predictive analytics using SageMaker and government datasets. Automated ETL and feature-engineering pipelines using AWS Glue, Redshift, and Step Functions, improving data refresh latency by 50%. Led development of policy analytics tools leveraging LLM based summarization and classification models tailored for public sector requirements. Developed a secure LLM driven internal policy chatbot, integrating it with Lambda, API Gateway, and S3 for versioning. Designed FAISS based semantic search engine enabling rapid retrieval of thousands of policy documents. Implemented model monitoring frameworks using MLflow, W&B, and SageMaker Model Monitor to track drift. Delivered data driven dashboards in Power BI to enhance transparency for state agencies. Keywords: continuous integration continuous deployment quality analyst artificial intelligence machine learning business intelligence sthree database information technology microsoft mississippi