Production LLM Systems  ·  Multi-Agent Orchestration  ·  Public-Sector AI

Agentic AI Architect
building scalable,
auditable AI systems.

20+ years of software engineering · 5+ years in AI/ML · 3+ years delivering enterprise LLM & agentic AI — focused on secure RAG, multi-agent workflows, and compliant production deployment.

View Projects Get in Touch
Mohammad Nasim, Ph.D.
Agentic AI RAG Systems Enterprise Arch. Ph.D. Research
01

About Me

Designing and delivering production-grade AI — compliant, auditable, and built to scale.

Agentic AI Architect and production LLM systems expert with 20+ years of software engineering experience, including 5+ years in AI/ML and 3+ years delivering enterprise-grade LLM applications. Proven expertise in designing RAG platforms, multi-agent systems, AI copilots, and compliance-oriented classification solutions using Python, FastAPI, Pydantic, Microsoft Agent Framework, Vertex AI, Gemini, FAISS, and cloud-native GCP services.

Strong domain experience across automotive, transportation, and insurance workflows, with a track record of building scalable, auditable, and production-ready AI systems for reasoning, retrieval, orchestration, and intelligent automation. Ph.D.-qualified technical leader with peer-reviewed publications and strong applied research depth, combining architecture, hands-on engineering, and business-aligned AI solution delivery.

0
+ years
Software Engineering
0
+ years
AI / ML Systems
0
+ years
Prod LLM / Agentic AI
Ph.D.
 
Computer Science Eng.
02

Core Technical Competencies

Full-stack AI delivery from model selection to production observability.

🤖

Agentic AI & LLM Systems

LangChain / LCEL LangGraph LangServe LangSmith RAG Systems Multi-Agent Systems Microsoft Agent Framework Google ADK MCP Protocol AutoGen / CrewAI Azure AI Foundry Claude Code / Co-work Prompt Engineering Tool Calling Structured Output Jinja Templates Gemini 1.5 / 2.5 OpenAI API Azure OpenAI Claude HuggingFace Transformers Self-hosted LLMs (LLaMA, Mistral, Vicuna)
☁️

Cloud & AI Infrastructure

GCP Vertex AI Vertex AI Embeddings Cloud Run / GKE BigQuery / GCS Dataflow / Dataproc Azure ML / Databricks Azure Data Factory AWS EC2 / Lambda / S3 SageMaker (exploratory) Docker / CI/CD Azure DevOps / Git / GitLab
⚡

API & Agent Development

Python FastAPI Flask / Django Pydantic Jinja Templates REST API Design API Integration Agent Orchestration Stateful Workflow Design Human-in-the-Loop Systems Streamlit
🔍

Vector Search & Knowledge Systems

FAISS Pinecone Weaviate ChromaDB Vertex AI Matching Engine Neo4j / GraphDB PostgreSQL / PostGIS SQLite / SQL Server
🧠

ML, NLP & Computer Vision

PyTorch / TensorFlow TensorFlow Lite HuggingFace Transformers OpenCV / YOLOv5 scikit-learn / LightGBM spaCy / NLTK Apache Spark / Kafka Spark Streaming MLflow / TensorBoard Pandas CVAT / LabelImg Ray RLlib
📊

Observability, Reporting & Governance

LangSmith Prompt Evaluation LLM Tracing Model Monitoring AI Governance Secure / Air-Gapped AI Power BI / Report Server Streamlit / JupyterHub ArcGIS / Mapbox SharePoint Jetson Nano
03

Professional Experience

Enterprise AI delivery across transportation, automotive, and insurance sectors.

AI Solution Architect

State of Rhode Island Department of Transportation via Osceola Staffing Holdings, LLC.
May 2024 – Present
  • Architected RAG knowledge assistant for vehicle technicians using Vertex AI Embeddings + FAISS with multi-step retrieval and reasoning via LangGraph; deployed via LangServe on Cloud Run.
  • Designed LLM-based TREAD complaint classification system with NHTSA regulatory reporting using Gemini 2.5 Flash/Pro, Google ADK, LangChain, and Pydantic for structured, auditable outputs.
  • Built LLM-powered customer support chatbot for vehicle issue triage using FastAPI, LangGraph, and Gemini 2.5 with Dataflow pipelines and BigQuery storage.
  • Developed multi-agent dental insurance support platform using Microsoft Agent Framework with central orchestrator and MCP-style specialized sub-agents.
  • Created computer vision pipelines for infrastructure defect detection (YOLOv5/OpenCV) and road sign cataloging (Jetson Nano/TensorFlow Lite) in air-gapped environments.
  • Built real-time highway incident detection system using Apache Kafka, Spark Streaming, and LLM-based alert generation confined within secure infrastructure.
  • Ensured AI governance compliance: data privacy, safety, auditability, and controlled data movement protocols throughout.
LangChain LangGraph Vertex AI Gemini 2.5 Google ADK MS Agent Framework FastAPI FAISS Cloud Run YOLOv5 Apache Kafka

AI Systems Architect

State of Michigan Department of Transportation via SystemDomain Inc.
Jun 2023 – Apr 2024
  • Built Neo4j knowledge graphs for bridge condition monitoring using HuggingFace Transformers on private GPU servers with Kafka + Spark Streaming pipelines.
  • Delivered NLP public feedback classifier using Azure OpenAI (private deployment) with spaCy/NLTK preprocessing and FAISS vector search in local containers.
  • Developed reinforcement learning snowplow route optimizer with Ray RLlib, Mapbox GL JS (offline maps), and PostGIS spatial data in a fully air-gapped deployment.
  • Prototyped multi-agent task delegation system with AutoGen + CrewAI for snow response planning — demonstrated ~15% improvement in task distribution efficiency.
  • Performed risk analysis and applied security checklists during AI deployment in public sector projects.
AutoGen CrewAI Neo4j Azure OpenAI Ray RLlib Apache Kafka HuggingFace PostGIS

Lead AI Developer

Atkins UK & SNC-Lavalin, Canada via Rainybits India Pvt. Ltd.
Apr 2021 – Aug 2022
  • Built predictive maintenance system on vehicle telemetry with Azure Data Factory + Databricks ETL and MLflow lifecycle tracking.
  • Deployed AI workforce planning assistants using HuggingFace Transformers with Pandas, spaCy, and Azure Databricks for distributed HR data processing.
  • Delivered Nextgen-GIS, Sweden Geo-portal, Lighting Scout App, and Digital Twin 3D framework for National Rail UK (IBM Maximo integration).
Azure Databricks HuggingFace scikit-learn LightGBM MLflow Power BI
Show earlier leadership roles (1998 – 2020) →
Senior Manager / ICT Expert
Public Authority of Electricity & Water (PAEW), Oman · Jan 2018 – Oct 2020
Managed 22-person planning & asset management team. Delivered two AI use-cases with TensorFlow, SSD, and YOLO for satellite building detection and parking intelligence. Created GIS/IoT/AI/ML systems with LiDAR, drone mapping, and deep learning workflows. Developed mobile GIS product in Android with RTK technology (±2cm accuracy).
Senior Manager – Projects
Cognizant (Fibre Inventory Management) · Mar 2016 – Aug 2017
Developed GIS Optical Fibre Network for Verizon Canada. Collaborated with Google Inc. on traffic annotation. Led 16 developers; deployed AngularJS + ArcGIS solutions on Azure SaaS.
Project Manager – Application Development
RMSI · Jul 2004 – Mar 2016
Delivered enterprise projects (e-governance portals, mobile GIS, web solutions) for clients across India, USA, Canada, Germany, Kenya, Qatar, Australia, Saudi Arabia. Notable: WAPCOS e-governance, Cloud Burst App (Tanzania), Water GIS (Qatar), World Bank SLB Connect.
Project Manager – Data Projects
Manchitra (Asianet GIS Mapping) · Dec 2002 – Jul 2004
Improved software team performance in MapInfo and GISNET. Developed automated tools for topological, spatial, and non-spatial data error checking.
Senior GIS Technologist / CAD Engineer
Cowi · INCA (deputed to TCS) · 1998 – 2002
International GIS enterprise databases, complex mapping, and AutoCAD customization with AutoLisp, Visual Lisp, and ObjectARX across UK, US, India, France, and El Salvador projects. Contributed to USWest Telecom fiber network digital development.
04

Featured Projects

Production AI systems across agentic architectures, RAG, compliance, and computer vision.

RAG Knowledge Assistant — Vehicle Technicians

RAG · GCP

Semantic retrieval over technical manuals, repair procedures, and service documents with Vertex AI Embeddings + FAISS. LangGraph orchestrates multi-step retrieval and reasoning; LangSmith provides full chain observability. Deployed via LangServe on Cloud Run.

Vertex AI Embeddings FAISS LangGraph Gemini 1.5 Flash LangSmith Cloud Run

LLM-Based TREAD Complaint Classification

Compliance · LLM

AI platform analyzing automotive conversations to assign NHTSA TREAD eligibility, primary, and additional defect codes. Multi-stage classification using Google ADK and Gemini 2.5. Generates structured, auditable outputs for quarterly regulatory reporting.

Gemini 2.5 Flash/Pro Google ADK LangGraph FastAPI Pydantic GCS

Customer Support Chatbot — Vehicle Issue Triage

LLM · NLP

LLM-powered chatbot capturing and interpreting customer-reported vehicle issues in natural language. Identifies issue patterns across battery, electrical, lights, paint/body damage, and warning indicators. Retriever-Generator pipeline with LangServe via Cloud Run.

Gemini 2.5 Google ADK LangGraph BigQuery Dataflow Cloud Run

Multi-Agent Dental Insurance Assistant

Multi-Agent · MCP

Central Orchestrator Agent coordinating specialized agents for Member Validation, Enrollment Inquiry, Benefit Lookup, and Provider Search. MCP-style backend tools integrated from enterprise data sources for natural-language insurance support.

MS Agent Framework MCP Protocol FastAPI Jinja Templates Pydantic

Multi-Agent Snow Response Planning

Agents · POC

Multi-agent AI prototype using AutoGen and CrewAI simulating logistics, weather, dispatch, and compliance agents. Agents used real weather feeds and local asset data. ~15% improvement in task distribution efficiency vs. prior year manual plans.

AutoGen CrewAI PostGIS Power BI Isolated VPC

Infrastructure Defect Detection from Video

CV · Air-Gapped

Computer vision pipeline detecting potholes and road defects from DOT surveillance feeds using custom YOLOv5 classifiers. CVAT annotation, TensorBoard tracking, fully air-gapped NAS training — no public cloud data transfer.

YOLOv5 OpenCV PyTorch CVAT TensorBoard

Real-Time Incident Detection & Alerting

Streaming · AI

AI system for real-time highway incident detection from video and sensor data. Apache Kafka + Spark Streaming for ingestion, LLM-based alert generation with LangChain + PyTorch. Deployed on internal SharePoint + Power Automate within secure infrastructure.

Apache Kafka Spark Streaming OpenCV LangChain SQL Server

AI-Powered Bridge Condition Monitoring

Graph AI · IoT

Knowledge graph system monitoring structural health of Michigan bridges using sensor and image data. Neo4j for inspection knowledge graphs; HuggingFace Transformers on private GPU servers; Kafka + Spark Streaming pipelines with JupyterHub collaboration.

Neo4j / GraphDB HuggingFace Apache Kafka Spark Streaming Azure AI Search

Autonomous Road Sign Detection

Edge AI · GIS

Vision system detecting and cataloging road signs from GPS-annotated dashcam footage. Jetson Nano devices with TensorFlow Lite for edge processing, ArcGIS Enterprise for visualization, YOLOv5 for detection. Fully disconnected from internet/cloud.

Jetson Nano TensorFlow Lite YOLOv5 ArcGIS SQLite

Predictive Maintenance for Fleet Vehicles

ML · Azure

Predictive system identifying vehicle maintenance needs from telemetry logs. ETL/ELT with Azure Data Factory + Databricks. ML models with scikit-learn, LightGBM, and MLflow tracking. Dashboards via Power BI Report Server — no external cloud exchange.

Azure Databricks scikit-learn LightGBM MLflow Power BI
05

Research & Recognition

Published researcher, hackathon finalist, and active open-source contributor.

Recent Publications

Adaptive Neural Audio Mixing with Human-in-the-Loop Feedback
AES International Conference 2025 · London, UK
AI-Powered Digital Human Clones
IBAC 2025 · New York, USA
Empirical Analysis of Functionalities and Confidence Scoring in Leading LLMs
IBAC 2024 (SUNY) · New York, USA
Fusion of Aerial Thermal Imaging with Deep Learning
IJITEE 2019 (SCOPUS) · Delhi, India
A Technology Centric Strategic Approach as Decision Support System During Flood Rescue
SEAHF 2019 (Springer) · Madrid, Spain
Finding Survivors in Flood Affected Areas by Deep Learning Approach
IJEAT 2019 (SCOPUS) · Delhi, India
Using Passive Anonymous Mobile Positioning Data & Aggregation Analytics for Flood Relief
IJEAT 2019 (SCOPUS) · Delhi, India

Awards & Recognition

🏆
Top 15% Finalist
Harvard D3 36-Hour Generative AI Earth Hackathon · 2024
🏅
Innovation Excellence Award
Advanced Automation System Development · 2013 & 2014
🏅
Technical Leadership Award
Enterprise Mobile GIS Implementation · 2012
⭐
RMSI Gold Star
Best Technical Talent · 2008

Education

Ph.D. in Computer Science Engineering
Lingaya's Vidyapeeth, India · 2020
Global M.S. in Business Analytics
Northwood University, USA · 2023
M.Sc. in GIS & Remote Sensing
TGOU, India · 2014
M.Tech in Computer Science & Engineering
SVU, India · 2013

Specialized Certifications

Agentic AI Fundamentals
Harvard D3 Institute
Certified AI Agent Developer
Vanderbilt University
Azure AI Fundamentals (AI-900)
Microsoft Certified
AI-Proficient Professional
Superhuman AI Academy
Career Essentials in Gen AI
Microsoft & LinkedIn
MS Certified Tech Specialist
.NET Framework · Web Apps

Community & Memberships

IEEE Member — AI/ML & Robotics Google Cloud Community LangChain Evangelist Open-Source Contributor Agentic AI n8n Contributor
06

Leadership & Mentoring

Building teams, setting technical standards, and growing the next generation of AI engineers.

8+
Engineers led in cross-functional agentic AI teams
15+
Junior developers mentored in AI/ML & agentic patterns
$2M+
Budget managed for AI infrastructure development
22
Team members managed at PAEW including engineers & QA
07

Get in Touch

Let's build the next generation of production-grade agentic AI systems together.

Contact Info

Available for new opportunities

Quick Message

Message is copied to your clipboard ready to paste into email.