Senior AI Solution Architect · AI/ML Engineer · LLM & Generative AI Specialist
15.5+ years in banking, analytics, and enterprise AI · 6+ years designing and delivering production LLM, Agentic AI, and computer vision systems

LinkedIn · Medium · er.rajkumaar@gmail.com

Senior AI Solution Architect and ML Engineer with 15.5+ years of experience delivering enterprise-grade AI systems across banking, finance, and language technology. I design end-to-end AI architectures — from data ingestion and model development to scalable cloud and on-prem deployment — with consistent focus on business impact, system reliability, and responsible AI.

Deep expertise in LLM-powered systems, Agentic AI workflows, Retrieval-Augmented Generation, and multilingual NLP. I have led globally distributed teams, driven AI modernization programs, and built production systems that are explainable, auditable, and maintainable at scale.

Currently architecting next-generation Agentic AI systems with multi-agent coordination, autonomous reasoning, and tool-augmented LLM pipelines.

Agentic AI & LLM Systems
Multi-agent orchestration, autonomous planning and reasoning, tool use, memory management, agent evaluation, LangGraph-based workflow design, and production-grade agentic pipeline architecture

LLMs & Generative AI
RAG pipelines, document QA systems, prompt engineering (zero-shot, few-shot, chain-of-thought), instruction tuning, multilingual LLM applications, Diffusion Models, DCGANs, ControlNet

Natural Language Processing
Transformer-based NER, semantic search, text classification, summarization, question answering, low-resource Indian language NLP, neural machine translation

Computer Vision
Object detection (YOLOv8), segmentation (SAM, SegFormer), OCR, real-time anomaly detection, generative vision pipelines, large vision models (Florence)

Machine Learning & Forecasting
Risk modeling, forecasting (ARIMA, LSTM, XGBoost), ensemble methods, Explainable AI (SHAP, LIME), full-lifecycle ML pipeline design

Solution Architecture & MLOps
End-to-end AI system design, Azure cloud deployment, Docker containerization, FastAPI and Flask microservices, CI/CD (GitHub Actions, GitLab), Hugging Face Hub, on-prem enterprise deployment, scalable API design

Analytics & Data Engineering
SQL, MySQL, DB2, A/B testing, hypothesis testing, advanced feature engineering, Power BI, Plotly, Seaborn

Production-style RAG chatbot with multi-document upload, image-aware retrieval, session memory, and Markdown chat UI — built with LangChain, ChromaDB, and Gemini. Deployed via Flask with Docker.

Real-time heat spike detection in thermal and surveillance video using OpenCV. Designed for industrial safety with alert logging and live visualization.

Vehicle speed estimation and over-speed alerting system built with YOLOv8 and OpenCV on real driving footage. Packaged as a containerized FastAPI service.

ML-based credit scoring for early-stage retailers with integrated Explainable AI (SHAP/LIME) for transparent, auditable lending decisions.

Neural machine translation for Angika — an endangered Indian language — using an encoder-decoder LSTM architecture in TensorFlow Keras.

Fine-tuned LLM for low-resource Angika language content generation and translation using Hugging Face Transformers. Contributes to digital preservation of an endangered language.

Modular pipeline for annotating driving scene videos with category-specific bounding boxes from CSV files — built for rapid prototyping and AI model support in video analytics.

• Explainable Sentiment Analysis for Financial News — ICIEM 2023
• Neural Machine Translation for Endangered Languages (Angika) — IIT Patna, 2023
• Generative AI for Drug Recommendations — SHADG 2024, IIT Kharagpur
• LLMs & DPT Models for Sentiment Analysis — NIT Patna, 2024
• Facial Image Encryption using DCGANs — M.Tech Thesis, IIT Jodhpur, 2023

• Led cross-functional AI/ML teams across geographies for complex enterprise delivery programs
• Acted as Agile Scrum Master — driving sprint velocity, team alignment, and measurable outcomes
• Mentored early-career engineers in Python, ML, NLP, and production deployment workflows
• Modernized legacy banking and analytics platforms by embedding scalable AI/ML pipelines
• Defined AI roadmaps and aligned data science initiatives to core business strategies
• Translated ambiguous business problems into structured, deliverable AI solutions

Certified SAFe® Scrum Master
Scaled Agile, Inc. · Issued March 2025 · Credential ID: 48725598-4060

Structured technical knowledge base covering Machine Learning, Deep Learning, NLP, Generative AI, RAG, Agentic AI, LLMOps, and Responsible AI — with concept guides, architectural patterns, and implementation references.

View Portfolio →

Open to collaborations and senior roles in AI Solution Architecture, Agentic AI systems, LLM engineering, and enterprise AI modernization.