Development Manager (AI)

I established and lead the firm's Generative AI function, owning AI strategy, target architecture, use-case delivery, hiring, and technical direction. I architected ten interoperating production Python and FastAPI microservices on Azure covering document intelligence, agentic retrieval, workflow automation, an LLM gateway, and machine-learning forecasting.

I delivered Azure AI Foundry agentic solutions using the Responses API, tool-calling loops, tiered model routing, and RAG over Azure AI Search. I also automated legal document processing using Azure Document Intelligence and LLM reasoning, and built case-volume forecasting models using XGBoost, LightGBM, statsmodels, and Optuna.

I own the AI Azure estate, CI/CD, Linux runtime stack, Docker, MongoDB, Redis, NGINX, Grafana, security remediation, and cost governance. I lead AI and web-development teams with over eight direct reports, manage the full employee lifecycle, author AI policy and operating frameworks, and coordinate delivery with UK leadership.

I delivered a production conversational AI platform for New India Assurance across web and WhatsApp channels, serving an estimated user base of 3.9 million. I developed Insurance GPT using a RAG pipeline, vector-database retrieval, cosine similarity, Ollama-served models, LangChain, QLoRA fine-tuning, and FastAPI deployment.

I also led multilingual chatbot implementations for government organisations, including electricity, water, and Regional Transport Office services, as well as MPSEDC and Unilever. These solutions used RASA NLU and dialogue management across REST API and WhatsApp channels.

I built custom machine-learning applications for prediction, automated reasoning, decision support, and optimisation. I also developed an NLP question-answering pipeline for intelligent information retrieval and response generation.

I designed, implemented, evaluated, and rapidly prototyped machine-learning and systems-engineering solutions. My responsibilities included end-to-end data transformation, feature preparation, model development, and experimentation.

Graduated with a GPA of 9.61 / 10.0.