Headline

Generative AI Engineer · Agentic Systems & LLM · RAG · LangGraph · FastAPI · Data Scientist | Published AI Researcher

About

I build AI systems that ship to production — not just models in notebooks. Over 3.5 years, I've designed Agentic RAG pipelines, fine-tuned LLMs (LoRA/QLoRA/GGUF), and delivered enterprise AI automation used in live recruitment workflows. My work reduced inference costs by 70% and processing latency by 30%. ━━ What I build ━━ Multi-agent orchestration systems (LangGraph, OpenAI, Qdrant) On-premise LLM deployments with guardrails & observability Secure AI APIs (FastAPI · JWT · AWS · GCP) NLP pipelines: NER, classification, resume parsing (95%+ accuracy) ━━ Tech stack ━━ Python · LangChain · LangGraph · OpenAI API · RAG · Qdrant · BM25 · FastAPI · SQL · AWS S3 · GCP · spaCy ━━ Research ━━ Published at the 4th International Conference on Data Science & Applications (Scopus-indexed, Springer). Topic: Deep Learning: Sentiment Analysis. ━━ Certifications ━━ Full Stack Data Science & AI (AlmaBetter) · Azure (Microsoft/SAP/Edunet) Currently focused on Generative AI, Responsible AI, and scalable LLM systems for enterprise use cases.

Location

Pune City

AI stack

antigravity

claude code

claude

codex

cursor

hugging face

Work experience

Data Scientist at Rangam, May 2024 - Present · 2 yrs 6 mos, Pune City, Maharashtra, India, • Architected Agentic RAG systems (LangGraph + OpenAI + Qdrant + BM25) for live recruitment automation — reduced retrieval latency by 30% with built-in guardrails and LLM-as-a-Judge evaluation. • Fine-tuned and deployed a 4B-parameter on-premise LLM using LoRA/QLoRA/GGUF quantization — cut AI inference costs by 70% vs cloud API approach while maintaining output quality. • Built multi-agent orchestration pipelines with zero-shot CoT reasoning, long-term memory, and full RAG observability (logging, tracing, evaluation) for explainable, production-grade outputs. • Delivered AI resume parser (LangChain + spaCy + AWS S3 + Google OCR) achieving 95%+ structured extraction accuracy across bulk ingestion pipelines. • Developed secure FastAPI microservices with JWT authentication and AWS/GCP infrastructure enforcing AI governance and compliance controls.

Data Scientist Trainee at AlmaBetter, Jan 2023 - Apr 2024 · 1 yr 4 mos, Bengaluru, Karnataka, India, • Built NLP pipelines for text classification and Named Entity Recognition using spaCy, NLTK, and Gensim — improved information extraction accuracy for downstream ML tasks. • Engineered features and conducted EDA on large retail datasets using Pandas; visualised findings with Matplotlib and Seaborn to guide modelling decisions. • Built and compared ML models (Linear Regression, Decision Trees, Random Forest) for retail sales forecasting — optimised Random Forest achieved 95% accuracy.

Education

Engineer's degree, Computer Engineering, Dr. Babasaheb Ambedkar Technological University

High School Diploma, N M Joshi VidhyaBhavan Goregaon