Open to opportunities

Saurav Kumar

@sauravkumar11

Message

LLM Engineer building production RAG systems and AI-native features that cut noise, cost, and latency.

India

Message

What I'm looking for

I want to build production RAG and AI-native products where I can design latency-aware pipelines, optimize cost, and ship measurable improvements like lower retrieval noise and higher engagement.

I specialize in production RAG systems and AI-native product development, focused on measurable outcomes. I built the core AI search and retrieval infrastructure for DocuSwift, indexing 500,000+ court judgments, and I also engineered a near-real-time personalization engine for a consumer book-tech platform.

I’m strong in LLM pipeline architecture (LangChain, LlamaIndex, LangGraph), hybrid vector search (HNSW/IVFFlat), and full-cycle Azure/AWS deployment. I ship systems that reduce retrieval noise, balance recall with query latency, and optimize cost—whether through semantic re-ranking, latency-aware design, or cost-optimized voice synthesis with ElevenLabs.

Experience

Work history, roles, and key accomplishments

Current

ML/LLM Engineer

Current

Kitab

Jul 2025 - Present (10 months)

Owned end-to-end design of a near-real-time recommendation engine in Supabase, using PostgreSQL triggers to update user taste embeddings and drive adaptive book recommendations. Built and benchmarked a multi-stage LLM summarization pipeline and implemented hybrid HNSW/IVFFlat vector indexing plus ElevenLabs TTS cost-optimization to reduce per-summary generation cost.

Supabase PostgreSQL HNSW IVFFlat SQL Triggers LangGraph Vector Search

Generative AI Developer

DocuSwift

May 2024 - Jun 2025 (1 year 1 month)

Led development of the Case Search™ RAG pipeline on Azure Cognitive Search, combining BM25 retrieval with semantic re-ranking to surface jurisdiction-relevant judgments from 500,000+ court cases with high precision and low latency. Architected Case Chat™ for natural-language Q&A over long case files and shipped zero-downtime serverless APIs with Azure Functions and Azure DevOps CI/CD.

Azure Cognitive Search RAG BM25 Prompt Engineering Azure Functions Azure DevOps fastAPI Chat History Session Management

Data Scientist

Soothsayer Analytics

Dec 2023 - May 2024 (5 months)

Designed and deployed machine learning models by tuning hyperparameters to improve prediction accuracy and computational efficiency. Conducted EDA with dashboards for non-technical stakeholders and monitored production models for drift, retraining proactively to maintain performance.

Model Training Hyperparameter Tuning scikit learn Dashboards Model Monitoring Monitoring For Drift SQL CI CD

Data Scientist

Agastya Data Solutions

Jun 2022 - Nov 2023 (1 year 5 months)

Developed quantitative investment strategies using ML on technical indicators and fundamentals, achieving 12–15% strategy returns. Optimized data processing to cut pipeline execution time by 50% and fine-tuned LLM prompts to improve reliability of downstream financial analysis outputs.

Quantitative Finance Machine Learning Feature Engineering Hyperparameter Tuning Data Visualization Python SQL