Open to opportunities

Sandeep Nayak

@sandeepnayak

Message

AI & LLM Engineer delivering production-ready RAG/agents with cost optimization and automation.

Japan

Message

What I'm looking for

I’m looking to build production LLM/RAG/agent systems end-to-end—focusing on reliability, observability, and measurable cost/value improvements. I want to work with teams that iterate fast, evaluate rigorously, and scale inference efficiently.

I’m an AI/LLM Engineer with 7+ years building production systems that turn business needs into measurable value—especially through RAG, LLM agents, and multi-agent workflows. At Gaudiy Inc, I delivered a BI conversational platform using Hybrid/GraphRAG over millions of user records, built multi-agent fan insight pipelines with sentiment and topic modeling, and reduced response time and compute cost with a multi-agent routing system.

I focus on end-to-end deployment, reliability, and observability: I co-architected an auto-scaling GPU inference platform on GKE with IaC (Terraform), CI/CD, and traceability (LangSmith + Datadog), and I introduced structured evaluation and prompt optimization using DSPy with guardrails. Previously at Rakuten Mobile, I improved analytics and extraction with LLM/RAG, led customer propensity and time-series models, and operationalized AI automation that cut ticket MTTR (patent applied); I also bring an operations-minded background from Senior Manager planning at Tata Motors.

Experience

Work history, roles, and key accomplishments

Current

AI Engineer

Current

Gaudiy Inc

Apr 2024 - Present (2 years 2 months)

Delivered a production BI conversational tool supporting a 10B yen investment story using hybrid RAG over millions of user records, and built multi-agent pipelines for fan sentiment, topic modeling, and merchandise extraction. Reduced debugging time ~30% with end-to-end observability, and co-architected a cost-efficient auto-scaling GPU inference platform on GKE.

Hybrid RAG Multi Agent Systems Langsmith DSPy GKE Kubernetes Terraform

Software Engineer (AI)

Rakuten Mobile

Sep 2022 - Mar 2024 (1 year 6 months)

Built an LLM analytics and knowledge-workflows stack using RAG-based extraction and summarization, cutting query handling time 40% and improving text processing efficiency 80%. Led customer propensity modeling, deployed a time-series DNN energy optimization model saving ~15M yen/year, and operationalized three production AI ticket systems (patent applied) that reduced manual effort by ~400 hours/p

RAG Hugging Face Pyspark Hadoop Time Series DNN Deep Learning Kafka Kubeflow MLOps

Software Engineer (AI)

Isha Foundation

Aug 2021 - Mar 2022 (7 months)

Built an entity resolution and recognition system that saved ~90 days/year/person by providing duplicate-entity recommendation and recognition using ML/NLP. Improved fuzzy-name-matching recall to 9.6% at 97.4% precision, and supported downstream workflow quality and volunteer user management via backend tooling.

Entity Resolution NLP Fuzzy Matching Duplicate Entity Recommendation Machine Learning Python Regex Django SQL

Senior Manager (Planning)

Tata Motors

Aug 2016 - Sep 2017 (1 year 1 month)

Reduced manufacturing process time by 40% using AI-driven value stream mapping to identify bottleneck parts in production workflows. Developed ML models for part-failure prediction to enable proactive maintenance and supported stakeholder decision-making with data-driven insights.

Python Value Streams Machine Learning Predictive Maintenance scikit learn Analytics Data Driven Decision Making

Education

Degrees, certifications, and relevant coursework

Tohoku University

Post-Graduate Researcher, Applied Machine Learning

2021 - 2022

Conducted applied machine learning research at Tohoku University, including 97% emotion estimation of working dogs from ECG and ML-based prediction of human motion for autonomous navigation.