Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

I'm Nikesh Bodduluri, an AI/ML Engineer with deep expertise in agentic LLM systems, retrieval-augmented reasoning, and large-scale machine learning. At OpenAI, I built production-grade agent frameworks for GPT-4o with multi-step planning, tool usage, memory persistence, and safety guardrails, deploying them with scalable Ray and Kubernetes infrastructure. I enjoy blending advanced LLM reasoning architectures with traditional ML systems to deliver reliable, explainable, and high-impact AI products. Previously at Citibank, I developed credit-risk and loan-default prediction models, architected Spark-based data pipelines, and implemented real-time scoring services aligned with IFRS 9 and Basel requirements. I’m passionate about building end-to-end AI solutions that are auditable, compliant, and capable of delivering real business value.…I'm Nikesh Bodduluri, an AI/ML Engineer with deep expertise in agentic LLM systems, retrieval-augmented reasoning, and large-scale machine learning. At OpenAI, I built production-grade agent frameworks for GPT-4o with multi-step planning, tool usage, memory persistence, and safety guardrails, deploying them with scalable Ray and Kubernetes infrastructure. I enjoy blending advanced LLM reasoning architectures with traditional ML systems to deliver reliable, explainable, and high-impact AI products. Previously at Citibank, I developed credit-risk and loan-default prediction models, architected Spark-based data pipelines, and implemented real-time scoring services aligned with IFRS 9 and Basel requirements. I’m passionate about building end-to-end AI solutions that are auditable, compliant, and capable of delivering real business value.

Nikesh Bodduluri

AI Developer, AI Engineer, Data Scientist, +3





I'm Nikesh Bodduluri, an AI/ML Engineer with deep expertise in agentic LLM systems, retrieval-augmented reasoning, and large-scale machine learning. At OpenAI, I built production-grade agent frameworks for GPT-4o with multi-step planning, tool usage, memory persistence, and safety guardrails, deploying them with scalable Ray and Kubernetes infrastructure. I enjoy blending advanced LLM reasoning architectures with traditional ML systems to deliver reliable, explainable, and high-impact AI products. Previously at Citibank, I developed credit-risk and loan-default prediction models, architected Spark-based data pipelines, and implemented real-time scoring services aligned with IFRS 9 and Basel requirements. I’m passionate about building end-to-end AI solutions that are auditable, compliant, and capable of delivering real business value.…I'm Nikesh Bodduluri, an AI/ML Engineer with deep expertise in agentic LLM systems, retrieval-augmented reasoning, and large-scale machine learning. At OpenAI, I built production-grade agent frameworks for GPT-4o with multi-step planning, tool usage, memory persistence, and safety guardrails, deploying them with scalable Ray and Kubernetes infrastructure. I enjoy blending advanced LLM reasoning architectures with traditional ML systems to deliver reliable, explainable, and high-impact AI products. Previously at Citibank, I developed credit-risk and loan-default prediction models, architected Spark-based data pipelines, and implemented real-time scoring services aligned with IFRS 9 and Basel requirements. I’m passionate about building end-to-end AI solutions that are auditable, compliant, and capable of delivering real business value.

Available to hire

I’m Nikesh Bodduluri, an AI/ML Engineer with deep expertise in agentic LLM systems, retrieval-augmented reasoning, and large-scale machine learning. At OpenAI, I built production-grade agent frameworks for GPT-4o with multi-step planning, tool usage, memory persistence, and safety guardrails, deploying them with scalable Ray and Kubernetes infrastructure. I enjoy blending advanced LLM reasoning architectures with traditional ML systems to deliver reliable, explainable, and high-impact AI products.

Previously at Citibank, I developed credit-risk and loan-default prediction models, architected Spark-based data pipelines, and implemented real-time scoring services aligned with IFRS 9 and Basel requirements. I’m passionate about building end-to-end AI solutions that are auditable, compliant, and capable of delivering real business value.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Intermediate

Intermediate

Language

English

Fluent

Work Experience

AI Engineer – Agentic LLM & Retrieval Systems at OpenAI

January 1, 2024 - Present

Designed and implemented agentic LLM frameworks for GPT-4o using state-machine orchestration, enabling multi-step reasoning, tool usage, and memory persistence. Built scalable Retrieval-Augmented Generation (RAG) pipelines with FAISS and ColBERT, reducing hallucinations and improving grounding. Developed function-calling and tool orchestration layers for dynamic API selection, and optimized long-context reasoning with hierarchical chunking for 200K+ token workflows. Implemented short-term and long-term agent memory with vector stores, deployed high-throughput inference on Ray Serve and Kubernetes, and integrated multimodal reasoning and safety guardrails. Created automated evaluation pipelines and collaborated with cross-functional teams to productionize capabilities.

AI/ML Engineer at OpenAI

January 1, 2024 - Present

Contributed to multimodal model training pipelines for GPT-4V and Sora, processing 500B+ token-equivalent image-text and video-text pairs using distributed data loading across OpenAI's A100/H100 GPU clusters. Built automated video-text alignment pipelines for data curation, implemented scene segmentation, caption quality filtering, and temporal consistency scoring that improved training data quality metrics by 29%. Developed vision-language evaluation frameworks with perceptual metrics (FID, CLIPScore, VQAv2 accuracy) and human preference protocols, reducing evaluation turnaround by 38%. Designed multimodal embedding alignment pipelines using contrastive learning, improving cross-modal retrieval by 26%. Contributed to Sora's video generation model inference optimization with temporal attention caching and mixed-precision serving, reducing per-video generation latency by 32% at production scale. Partnered with Safety and Alignment teams to integrate multimodal content classifiers, reduc

Machine Learning Engineer – Credit Risk & Default Prediction at Citibank

March 1, 2018 - June 1, 2022

Developed credit risk and loan default prediction models (XGBoost, LightGBM), reducing default losses by 18%. Built large-scale Spark/PySpark data pipelines processing millions of borrower records. Designed PD, LGD, and EAD models for IFRS 9 ECL calculations. Implemented feature engineering capturing borrower behavior trends, improving model AUC. Established time-based validation frameworks, automated ML workflows with Airflow and MLflow, and integrated real-time scoring services via REST APIs. Developed SHAP-based explainability for regulatory audits and built dashboards to monitor model performance and drift, ensuring Basel III compliance.

ML Engineer at Adobe (Experience Cloud) - India

March 1, 2018 - June 1, 2022

Built productionized propensity scoring models using gradient-boosted trees (XGBoost/LightGBM) and deep learning for 100M+ profiles (purchase intent, churn, upsell). Developed real-time audience segmentation pipelines with Apache Spark and Adobe Experience Platform, delivering sub-hourly segment refreshes. Created content recommendation models with collaborative filtering and neural embeddings, achieving ~27% uplift in content CTR in Adobe Target AB tests. Designed multi-touch attribution models to optimize marketing spend, delivering ~20% ROI uplift. Implemented real-time personalization scoring with Kafka and Redis, achieving sub-25ms p95 latency in a multi-tenant Sensei environment. Migrated batch features to streaming on Azure Event Hubs, reducing feature staleness from 24h to 45m. Built automated model monitoring and drift detection across 20+ features, cutting time to detect degradation by ~43%. Standardized offline (AUC-ROC, PR) and online (conversion lift, revenue per visitor)

Machine Learning Engineer at Adobe

March 1, 2018 - June 1, 2022

Built productionized customer propensity scoring models using gradient boosted trees (XGBoost/LightGBM) and deep learning architectures to predict purchase intent, churn risk, and upsell likelihood across 100M+ profiles. Developed real-time audience segmentation pipelines with Apache Spark and Adobe Experience Platform (AEP), refreshing segments sub-hourly. Built content recommendation models using collaborative filtering and neural embeddings, improving content click-through rates by ~27% in controlled AB tests on Adobe Target. Designed multi-touch attribution frameworks identifying high-value conversion pathways, enabling a 20% uplift in attributed ROI. Implemented real-time personalization scoring with Apache Kafka and Redis, achieving sub-25ms p95 inference latency in a multi-tenant architecture. Migrated batch feature pipelines to streaming on Azure Event Hubs, reducing feature staleness from 24 hours to 45 minutes. Built automated model monitoring and data drift detection across