{Ismail
Shaikh}

AI Solution Specialist / Agentic AI / AI Solution Architecture

I design and ship AI agents that read, decide and act, and take them from business requirements to production with guardrails, observability and cost in mind. Twelve years across embedded systems, RPA and Python backends; today I lead AI delivery for an enterprise team.

View my work Resume
vuln-triage-agent
$ agent run vuln-triage --source cert-inbox
[ingest] 38 new advisories · email, GitHub, RSS
[extract] LLM → CVE, product, version, severity ✓ 38/38
[judge] LLM-as-Judge verified 35 · 3 flagged
[review] 3 sent to human-in-the-loop queue
[match] matched to customer assets → 9 relevant
[route] 9 advisories sent to 4 owning teams
✔ done in 54s · manual process: ~2 h a day
Simulated replay with illustrative numbers. No AI calls, no tokens.How it works ↓
~/work

AI systems running in production, and the automation work behind them.

In productionEnterprise · details anonymised

AI Vulnerability Automation Platform

Daily advisory triage: 2 hours → about a minute

Security advisories arrive all day by email, GitHub and RSS. Someone used to read every one, work out which customers were affected and forward it to the right team. I designed and built an agentic platform that does this end to end, with an AI check and a human review step before anything is routed.

PythonDjangoLLMsLLM-as-JudgeHuman-in-the-loop
  1. 1
    IngestCERT advisories from email, GitHub and RSS feeds
  2. 2
    ExtractAn LLM pulls out CVE, product, version and severity
  3. 3
    JudgeA second LLM checks each extraction before it is trusted
  4. 4
    Human reviewUncertain cases go to an analyst instead of straight through
  5. 5
    MatchCompared with each customer's asset inventory
  6. 6
    RouteSent to the team that owns the affected systems
In productionEnterprise · details anonymised

Enterprise Document RAG Solution

Answers from technical documents, plus first-draft HLD/LLD designs

Teams needed answers buried in long technical PDFs, many with tables that run across several pages. I designed and built a retrieval-augmented generation system that handles those tables, answers questions with sources, and drafts high- and low-level design documents. I also designed its LLM hosting and weighed cost against performance.

LangChainLlamaIndexCustom chunkingEmbeddingsVector database
  1. 1
    ParsePDFs, including tables split across pages
  2. 2
    ChunkCustom chunking keeps table rows and their headers together
  3. 3
    Embed & indexStored in a vector database
  4. 4
    RetrieveThe most relevant passages for each question
  5. 5
    GenerateGrounded answers and first-draft HLD/LLD documents
Open sourceOngoing

Full-Stack Document RAG Platform

A self-hosted RAG stack, from parsing to tracing

A reference architecture I build and run myself: document parsing, retrieval, agents, local or remote models, LLM tracing and a chat interface, all running in containers.

DoclingLlamaIndexCrewAIPostgreSQLOllamaArize PhoenixOpenWebUIDocker
View on GitHub →
  1. 1
    DoclingTurns documents into clean, structured text
  2. 2
    LlamaIndex + PostgreSQLIndexing and retrieval
  3. 3
    CrewAI agentsPlan and answer multi-step questions
  4. 4
    OllamaLocal and remote models
  5. 5
    Arize PhoenixTraces every call for evaluation and debugging
  6. 6
    OpenWebUIThe chat interface

Automation delivery (RPA), 2018 to 2022

25+ automations · teams of 4–6 · ~400 hrs saved a month
Manufacturer of scales and analytical instruments

24x7 alarm-handling bot

A Python middleware service pulls new alarms from Cisco into MySQL, and an Automation Anywhere bot picks them up and processes them around the clock. Delivered the full life cycle solo.

AA 360PythonMySQLTask Scheduler
Food & beverage company, Singapore

Self-service password reset

Users raise a reset request through Power Apps or a chatbot, and a bot resets the password in the application without IT stepping in.

Power Automate cloud flowsDesktop flowsPower Apps
Food & beverage company, Singapore

Weekly web report automation

A bot downloads files from a website, cleans and reshapes the data in Excel and sends the finished report to the customer every week.

AA 360Excel macrosPower Query
Food & beverage company, Singapore

Multi-site report collection

Downloads reports from several websites, merges and cleans them in Excel and emails the result. Built end to end on my own.

Power Automate DesktopCloud flowsPower Query
  • Monthly margin billingGenerates monthly billing with Excel invoices.
  • V11 to A360 migrationMoved a group of bots from Automation Anywhere V11 to A360.
  • Oracle PO updatesUpdates purchase orders in the Oracle desktop app.
  • Incident creationRaises incidents automatically from monitoring alerts.
  • Role provisioningGrants and updates user roles in business applications.
  • Service contract renewalHandles renewal of service contracts end to end.
  • Insurance cancellationProcesses customer requests to cancel insurance policies.
  • Insurance bookingBooks policies and creates customer profiles.
  • MIS reportingDownloads raw data, builds a custom report and sends it to the business.
  • Price scrapingCollects insurance product prices from competitor websites for marketing.
  • Promotion card generationProduces daily promotion cards for retail stores.

Roofing ERP

Customers, site measurements, quotations, material stock, job scheduling and invoicing in one system, replacing spreadsheets and paper job sheets.

Django · MySQL

Electronics Inventory ERP

Components by part number and location, stock in and out, low-stock alerts, suppliers, purchase orders and stock-value reports.

Django · MySQL
~/approach

How I take an AI use case from idea to production.

  1. 01

    Assess

    Sit with the people doing the work, map every step and pick where AI or automation saves the most time.

  2. 02

    Design

    Choose the architecture, plan for messy data and decide where an AI check or a human review is needed. Estimate the running cost.

  3. 03

    Build

    Agents, retrieval and APIs, with tracing from day one so every model call can be inspected.

  4. 04

    Run

    Test on real cases, document the solution, hand over, and keep improving it as the process changes.

~/experience

Twelve years, from circuit boards to AI agents.

Embedded & IoT→RPA & automation→Python backends→AI agents

  1. 10/2022 – PresentMumbai

    AI Solution Specialist

    Orange Business Services

    • Team lead: lead 8-10 engineers across AI and backend projects, owning technical design and delivery.
    • AI Vulnerability Automation Platform: designed and built; cut daily triage from 2 hours to about a minute, saving ~40 hours a month.
    • Enterprise Document RAG Solution: designed and built, including LLM hosting and cost-performance design.
    • Backend and data: API design and data modelling for Django REST services; Python ETL pipelines.
  2. 01/2019 – 10/2022Mumbai

    RPA Development Consultant

    Infogain India

    • Led a team of 4–6 developers on client automation, from process discovery to production support.
    • Delivered 25+ automations, saving an estimated 400+ hours of manual work a month.
    • Wrote the solution and process design documents used for client sign-off.
  3. 07/2018 – 11/2018

    RPA Developer

    XLNC Technologies

  4. 02/2016 – 05/2018

    Senior Electronics Engineer

    Kinikar Electronics & Tech Services

  5. 09/2014 – 01/2016

    Junior Engineer

    Ilogic Academy

~/skills

What I build with.

AI architecture & delivery

AI solution architectureSystem designRequirements to productionCost & scalability trade-offsTeam leadership

GenAI & agents

LangChainLangGraphCrewAILlamaIndexRAGMulti-agent systemsMCPPrompt engineeringOpenAI APIOllama

Responsible AI & LLMOps

LLM-as-JudgeHuman-in-the-loopArize PhoenixLLM tracing

Retrieval & document AI

DoclingEmbeddingsVector databases

Backend & platform

PythonDjangoDjango REST frameworkFastAPIFlaskREST APIsMicroservicesDocker

Databases

PostgreSQLMySQLMS SQL ServerMongoDBSupabase

Automation

n8nAutomation Anywhere 360UiPathPower AutomatePower AppsVBAPower Query

Embedded & IoT

Embedded CMicrocontrollersModbusI2C
~/certifications

Awards, certifications and education.

Bot Games Season 4, Trafica award

Automation Anywhere · November 2023

Databricks AcademyGenerative AI

  • Generative AI Fundamentals
  • Generative AI Application Evaluation and Governance
  • Generative AI Application Deployment and Monitoring
  • Building Retrieval Agents on Databricks
  • Building Single-Agent Applications on Databricks

Agents & LLM toolingAnthropic, LangChain, Hugging Face

  • Introduction to Model Context ProtocolAnthropic Academy
  • Claude Code in ActionAnthropic Academy
  • Introduction to LangChain (Python)LangChain Academy
  • Introduction to Deep AgentsLangChain Academy · in progress
  • LangChain Certified Agent EngineerLangChain · in progress
  • AI Agents CourseHugging Face · in progress

AutomationRPA platforms

  • Certified Advanced RPA Professional (Automation 360)Automation Anywhere · 2023
  • Document Automation MasterclassAutomation Anywhere · 2024
  • RPA Developer Foundation DiplomaUiPath

Python & backend

  • REST API with Django REST Framework and Docker (TDD)
  • CS50's Introduction to Programming with PythonHarvard · 2023
  • Python for Data Science, AI & DevelopmentIBM on Coursera
  1. 2026Master's degree

    MSc, Computer Science

    University of Mumbai

  2. 2014Bachelor's degree

    BE, Electronics & Telecommunications

    NES's Gangamai College of Engineering

~/resume

My resume, always the latest version.

Resume of Mohammed Ismail Shaikh, AI Solution Specialist: profile, experience, skills, certifications and education

Mohammed Ismail Shaikh

AI Solution Specialist · AI Solution Architecture. PDF.

Download resume (PDF) Open full size Plain-text version for job portals
Last updatedOctober 2026
Notice period60 days
Experience12 years
~/contact

Building something with AI agents? Let's talk.

hello@shaikhismail.com
LinkedIn GitHub Blog Free tools