Open to opportunities

Saran Puneet

@saranpuneet

Data & AI Platform Engineer building production RAG and streaming pipelines, bridging distributed systems with LLM research.

India

Message

What I'm looking for

I’m looking to build production GenAI/RAG systems with streaming data platforms and multi-agent workflows—engineering low-latency, reliable pipelines, strong evaluation (citations/faithfulness), and practical autonomy that teams can trust.

I’m a Data & AI Platform Engineer who builds production GenAI systems end-to-end, from streaming data pipelines to enterprise RAG experiences. At Ola Electric, I engineered telemetry ingestion processing “1 TB/day” of streaming EV sensor data and shipped robust testing outcomes while coordinating cloud migration work.

In my current role as Founder & AI/Data Engineer at Webbverse AI, I architect enterprise-grade GenAI platforms that integrate streaming data (Kafka, Spark, Databricks) with multi-agent LLM workflows. I’ve also focused on engineering autonomous error-recovery pipelines and serverless deployments using NVIDIA NIM and LiteRT to reduce LLM inference latency—while supporting mission-critical enhancements through confidential defense-sector engagements.

Previously at Kindrix Group, I created scalable AI data platforms for early-stage startups, integrating RAG pipelines using LangChain and vector database components for document parsing and retrieval. I owned RAG system performance—precision, source citation, and hallucination control—so teams could reduce enterprise document review cycles and deliver dependable outputs.

I began strengthening my foundations through ETL/ELT pipeline delivery at Sapper AI (Zyient), including a data mapping framework that standardized heterogeneous JSON/text flows into structured formats for downstream AI processing. Earlier, I built and published a novel NER data augmentation pipeline for Samsung Bixby wake-word detection, and I continue to engineer fast by pairing hands-on systems work with modern AI-augmented development tooling.

Experience

Work history, roles, and key accomplishments

Current

Founder & AI/Data Engineer

Current

Webbverse AI

Feb 2026 - Present (5 months)

Founded a specialized AI consultancy architecting enterprise-grade GenAI systems that integrate streaming data platforms (Kafka, Spark, Databricks) with multi-agent LLM workflows for B2B solutions. Engineered autonomous error-recovery pipelines and serverless deployments (NVIDIA NIM, LiteRT) to cut LLM inference latency, and collaborated with the Indian Army (Western Command) on surveillance-relat

Kafka Spark Databricks Gen AI Systems RAG AI Agentic Workflow Vector Search NVIDIA NIM Self Healing Pipelines Serverless

Data Engineer

Kindrix Group

Jun 2024 - Dec 2025 (1 year 6 months)

Created scalable AI data platforms for early-stage startups by integrating RAG pipelines for enterprise document parsing and retrieval. Owned end-to-end RAG performance (retrieval precision, source citation, and hallucination control), reducing document review cycles for enterprise clients.

RAG Vector Databases Retrieval Precision Source Citation

DPE (Backend)

Sapper AI (Zyient)

Jul 2023 - Nov 2023 (4 months)

Delivered automated ETL/ELT pipelines feeding LLM-powered RAG systems, reducing manual review effort for three enterprise clients. Designed a robust data mapping framework to standardize heterogeneous JSON/text flows into structured formats, cutting unstructured data ingestion time by 40%.

ETL ELT Data Mapping JSON Data Ingestion Data Standardization Automated pipelines

SDE I

Ola Electric

Jun 2022 - Jul 2023 (1 year 1 month)

Engineered telemetry ingestion pipelines processing 1 TB/day of streaming EV sensor data and achieved 85% unit and integration test coverage. Coordinated a cloud migration from on-premise to AWS/Azure, reducing data retrieval latency by 30% and infrastructure cost by 25%.

Telemetry AWS Azure Cloud Migration Unit Testing Integration Testing

PRISM Developer

Samsung

Jul 2020 - Apr 2021 (9 months)

Engineered a NER data augmentation pipeline using Word2Vec, ELMo, BERT, Seq2Seq grammar correction, and a cosine-similarity sanity checker for Samsung Bixby wake-word detection. Improved F1 on CoNLL-2003 and published the work in the International Journal of Speech Technology (Springer, 2023).

Word2Vec ELMO BERT Cosine Similarity RoBERTa NLP Wake Word