Pravin Pawar
Pune
Summary
Lead Data Scientist and Gen AI Product Architect with 11 years of experience evolving from NLP and machine learning to designing enterprise-scale agentic AI platforms. Architected multi-agent, RAG-based systems integrated across products, serving 3K+ users and delivering measurable productivity gains. Expertise in AI platform architecture, agent orchestration, and scalable LLM systems, with strong ownership of cross-team adoption and system evolution.
Overview
10
10
years of professional experience
Work History
Lead Data Scientist | Gen AI Product Architect
Johnson controls
Pune
09.2023 - Current
- Architected and led development of a multi-agent Generative AI platform integrated across 3+ enterprise applications (web & mobile), serving 3K+ MAU and improving field technician productivity by 1-2 hours per day.
- Designed a modular agent orchestration framework (agent schemas, I/O contracts, planning-based execution), enabling scalable integration across multiple teams and use cases.
- Built a multi-domain RAG architecture (HVAC, Controls, Fire) with hybrid retrieval, re-ranking, and grounding strategies to ensure accurate, context-aware responses across heterogeneous data sources.
- Defined and implemented LLM routing strategy across models for reasoning, orchestration, and response generation, optimizing latency, cost, and quality trade-offs.
- Led cross-functional platform adoption; mentored and managed a team of 9 engineers/data scientists, driving architecture decisions impacting multiple products.
- Designed scalable ingestion and indexing pipelines with automated scheduling for continuous updates across structured and unstructured enterprise data sources.
- Improved system efficiency through caching and retrieval optimizations, maintaining 9-11s response latency while reducing unnecessary LLM calls.
- Defined platform evolution roadmap toward a unified AI field assistant, including agent containerization, modular extensibility, and multimodal capabilities (voice + text).
AI / Engineering Lead
Wipro
Pune
01.2022 - 09.2023
- Led design and delivery of LLM-powered and NLP-based enterprise solutions across domains including legal, medical, and customer intelligence.
- Architected and executed AI solutions for global clients such as Johnson & Johnson, Huawei Technologies, and Integra LifeSciences.
- Built and led a team of 6 engineers, delivering multilingual classification and keyword extraction systems with ~91% performance accuracy.
- Drove end-to-end pipeline redesign and MLOps improvements, enhancing scalability, maintainability, and deployment efficiency.
- Developed POVs on Synthetic Data Generation (SDG) and early-stage LLM applications, contributing to AI capability building.
- Recognized as top performer for driving engineering best practices and delivering high-impact AI solutions.
Senior Project Engineer
Wipro
Pune
02.2019 - 12.2021
- Achieved 91% F1 score in production-grade classification systems, improving significantly from POC benchmarks.
- Led end-to-end system integration, coordinating across teams to deliver scalable and production-ready AI solutions.
- Delivered enterprise AI solutions for clients including Huawei Technologies, Hershey's, and Refinitiv.
- Designed and developed large-scale NLP and document intelligence systems, including classification, keyword extraction, and information retrieval pipelines.
- Built optimization modules (deduplication, ranking using MinHash/SimHash), improving efficiency and model performance.
- Processed 750K+ documents in large-scale document intelligence pipeline for enterprise use cases.
- Developed AI-powered legal document review system; filed patent for novel information extraction and feedback learning approach.
Project Engineer
CDAC
Pune
11.2015 - 01.2019
- Built NLP and computer vision solutions for government-scale applications with Ministry of Electronics and Information Technology and Election Commission of India.
- Developed voter ID extraction and deduplication system with OpenCV and FaceNet-based embeddings, improving accuracy and efficiency in voter identification.
- Designed large-scale multilingual word embeddings (Hindi, Marathi, English) utilizing datasets exceeding 1B tokens, facilitating cross-language applications.
- Implemented RNN-based language models for autocorrect and autosuggest, improving OCR/OHWR accuracy by 3-4%.
- Built post-processing pipelines using trie-based and statistical NLP techniques, significantly enhancing overall system performance and output quality.
Education
Executive Program in Business Management -
IIM Indore
Indore10-2026
MTech - Signal Processing
College of Engineering Pune
Pune07-2015
BTech - Electronics and Telecommunication
Dr. Babasaheb Ambedkar Technological University, Lonere
Raigad, Maharashtra, India07-2012
Skills
- AI Platform Architecture
- Multi-Agent Systems
- Agent Orchestration
- Scalable System Design
- Modular Architecture
- API Design
- Microservices
- Generative AI
- Large Language Models (LLMs)
- Retrieval-Augmented Generation (RAG)
- LLM Orchestration
- LLM Routing
- Prompt Engineering
- LangChain
- LangGraph
- LlamaIndex
- Semantic Kernel
- TensorFlow
- PyTorch
- Scikit-learn
- Azure OpenAI
- GPT-4/5
- LLaMA
- Mistral
- Vertex AI
- Amazon Bedrock
- Pinecone
- FAISS
- Azure AI Search
- Milvus
- Hybrid Search
- Semantic Search
- Re-ranking
- Classification
- NER
- Summarization
- Sentiment Analysis
- Clustering
- MLflow
- Azure AI
- AWS SageMaker
- GCP Vertex AI
- Python
Languages
English
Proficient (C2)
C2
Hindi
Proficient (C2)
C2
Marathi
Native
Native
Timeline
Lead Data Scientist | Gen AI Product Architect
Johnson controls
09.2023 - Current
AI / Engineering Lead
Wipro
01.2022 - 09.2023
Senior Project Engineer
Wipro
02.2019 - 12.2021
Project Engineer
CDAC
11.2015 - 01.2019
Executive Program in Business Management -
IIM Indore
MTech - Signal Processing
College of Engineering Pune
BTech - Electronics and Telecommunication
Dr. Babasaheb Ambedkar Technological University, Lonere