AI Engineer · Machine Learning · OSS Contributor

I build scalable AI solutions with a strong foundation in LLMs, agentic systems, and full-stack development. Currently working at HGS Global Solutions since April 2025.

📫 nimashiri2012@gmail.com  ·  LinkedIn  ·  Google Scholar

• LLM & agentic systems — building production RAG pipelines, multi-agent workflows, and LLM-powered applications
• AI infrastructure — self-hosted inference, cloud deployment, observability, and evaluation
• Full-stack development — REST APIs, web apps, and databases end-to-end

AI-powered Continuing Healthcare checklist platform built for UK local authorities

• Reduced CHC checklist drafting time by 80%+ (from ~4 hours to ~30 minutes)
• Achieved zero ICB rejections in pilot (previously 1 in 10)
• Shortlisted for 9 sector awards — MJ Awards, LGC Awards, HTN Awards
• HIPAA/GDPR-compliant architecture; UK-hosted infrastructure via HGS UK

Orchestrator-led platform for building AI assistants and digital workers — HGS Global Solutions

• Architected a multi-tenant platform supporting pluggable cloud and self-hosted LLM runtimes
• Deployed digital workers resolving customer webforms at 900+ requests/day with minimal human-agent intervention
• Integrated a call-center chatbot for inbound customer queries and automated auto-finance approval workflows
• Built with LangGraph, LangChain, end-to-end Opik tracing for observability, and dual deployment models (dedicated client infra or agentic SaaS)

Self-hosted, cost-effective alternative to cloud LLM APIs — HGS Global Solutions

• Engineered a private on-premises LLM inference platform using vLLM on Amazon EC2
• Enables enterprise clients to run open-weight models without relying on third-party cloud APIs
• Supports the agentic platform as a pluggable self-hosted LLM runtime

RAG-powered web tool for auto-drafting enterprise RFP responses — HGS Global Solutions

• Built with Django, Celery, LlamaIndex, LlamaParse, and Azure AI Search
• Parses uploaded RFPs and auto-drafts section-level responses — executive summary, legal terms, disqualifiers, deliverables, and Q&A
• Grounds responses in enterprise corpora via reranked retrieval with source citations, exported as structured Word documents

AWS

Azure

DevOps

This profile was crafted with the help of Claude by Anthropic.