PRINCIPAL SOLUTIONS ARCHITECT

AkhileshMishra

I help enterprises turn AI prototypes into governed, measurable business systems.

12 years across automation, cloud, compliance, and GenAI — with proven savings, adoption, and delivery metrics.

WhatsAppDownload CV
Akhilesh Mishra

Verified Professional Certifications

Microsoft Certified: Agentic AI Business Solutions ArchitectMicrosoft Certified: Agentic AI Business Solutions ArchitectVerify →AWS Certified Generative AI Developer – Professional Early AdopterAWS Certified Generative AI Developer – Professional Early AdopterVerify →AWS Certified Security – SpecialtyAWS Certified Security – SpecialtyVerify →AWS Certified Solutions Architect – ProfessionalAWS Certified Solutions Architect – ProfessionalVerify →Microsoft Certified: Azure AI Engineer AssociateMicrosoft Certified: Azure AI Engineer AssociateVerify →

12 Years.
4 Transformations.

I've been at the front line of every major enterprise technology shift since 2012 — not as a spectator, but as the architect delivering the solution. Each wave taught me that technology alone doesn't transform businesses; the right architecture, at the right friction point, with measurable outcomes does.

012012

Web & CMS

WordPress, Joomla, VB script apps, MVC cloud applications. Full-stack freelance developer.

022016

RPA

Blue Prism, UiPath, Kofax at EY, Deloitte, Shell. Enterprise automation across finance and tax.

032022

Agile at Scale

39-person team at ASM. 95% SLA, 40% org-wide adoption, COE governance for semiconductors.

042024

Gen AI

Enterprise GenAI platforms, Agentic AI, RAG, multi-model systems, AI-native compliance.

Beyond AI

Non-AI Experience

AI is my current focus, but it's built on a deep foundation of enterprise delivery, automation, cloud infrastructure, and team leadership.

⚙️

RPA & Automation Delivery

UiPath, Blue Prism, Kofax, Python, Selenium, C#, Django REST APIs. 60% manual processing reduction. $2M+ programmes managed.

☁️

Cloud & DevSecOps

AWS, Azure, Government Cloud (GCC), IaC, Terraform, VAPT, SAST, CI/CD pipelines. Led DevSecOps teams with automated security scanning.

📊

Enterprise Portals & Workflows

Airol Portal: single dashboard for costing, POs, stock, dispatch. 8/8 pain points mapped. 98.7% cycle-time reduction on cost updates.

🔗

SAP & ERP Integration

SAP MM/GTS decoupling, SharePoint landing zones, Power Automate orchestration, Power BI visibility, Vendor Master write-back.

👥

Leadership & Agile Delivery

Teams up to 39 professionals. 95% SLA maintenance. Agile methodologies, sprint governance, COE frameworks, change management.

🎯

Consulting & Process Design

Engagement lifecycles, strategy diagnostics, business-case development, POC delivery for RFPs, client relationship management.

Three Problems.
Three Original Solutions.

Each began with a business problem that traditional approaches couldn't solve.

01

The Supplier Who Wouldn't Log In

P2P Compliance · Multi-Channel AI Ingestion

Tier-2 suppliers in China don't use enterprise portals. They use WeChat. So instead of forcing them into our world, we brought compliance to theirs.

Approach

Decoupled SAP via SharePoint, triggered Power Automate, reached vendors on WhatsApp/WeChat, validated with AI Builder OCR, routed low-confidence through Dataverse HITL, wrote back to SAP.

Tech Stack

SAP MM/GTSPower AutomateAI BuilderTwilioWeChat APIn8nPower BIDataverse

Role

System Owner &
Solution Engineer

97.4%

Compliance

from 62%

<24h

Cycle Time

from 7-14 days

0%

Demurrage

from frequent

Real-Time

Visibility

from manual

Proof:Live production system at Keppel • SAP Vendor Master records updated • Power BI dashboard active
02

The Auditor Who Couldn't Keep Up

SOLAR · Enterprise Compliance Intelligence

What if we taught AI to read policy documents, query logs with generated SQL, and write the gap report itself?

Approach

AI compliance platform on AWS: Bedrock Claude reasons over policy, generates Athena SQL, RAG KB provides context, Step Functions orchestrate parallel analysis, automated reports with gap identification.

Tech Stack

Bedrock ClaudeAthenaGlueStep FunctionsLambdaDynamoDBRAG KBTerraform

Role

AI Automation
Architect

119+

Commits

shipped

4

States

compliance

Multi

Sources

CSV/XLSX/PDF

Auto

Gap Detection

+ remediation

Proof:Open-source on GitHub • 119+ commits • Terraform IaC • React dashboard • CI/CD deployed
03

The Agent That Validates Itself

AI-Native Validation Gates · Regulated Agentic Software

Validation can't be an event anymore. It has to be a property of the pipeline itself.

Approach

Five automated gate families in CI/CD: static checks, agentic behavior evals, boundary security, production evidence/approval, continuous runtime revalidation. Demonstrated with pharma HCP Brief Generator.

Tech Stack

GitHub ActionsLangfuseLiteLLMBedrockOPA/CedarEventBridgeKMSGarak

Role

Section Owner
Gates & Deploy

5

Gate Families

lifecycle

300

Scenarios

golden set

10

Artifacts

evidence pack

24/7

Monitoring

runtime

Proof:Live workshop demo delivered • Handoff contract signed • Validation deck presented to stakeholders

The Strategy

The AI Flywheel

I implemented a self-funding AI adoption model: use AI to accelerate tech delivery, eliminate legacy license costs, then reinvest the savings into high-value business AI use cases.

1

Fund the Engine

AI for Tech Teams

Deployed AI dev agents (Kiro) with validation gates to ship traditional business apps — workflow automations, RPA replacements — as native serverless code. Eliminated legacy license-based tooling entirely.

~SGD 36K/yr

Legacy RPA & automation licenses eliminated

Replaced with serverless at ~93% cost reduction

2

Scale to Business AI

Deal Evaluation for Assets

With credibility and budget from Phase 1, funded a high-value business AI use case: automated Deal Evaluation for asset acquisition. Keppel, as an asset management company, evaluates investment opportunities at scale.

Architecture

• Multi-model deep research pipeline (market data, financials, risk factors)

• LLM-as-a-Judge for cross-validation and consensus scoring

• Final AI assessment layered on top of Investment Memo

• Human-in-the-loop for final investment committee decision

Bedrock ClaudeMulti-ModelRAGLLM-as-JudgeStep FunctionsS3OpenSearch

"Use AI to save money on tech delivery. Use the savings to fund AI that makes money for the business. Repeat."

The Numbers Don't Lie

0%

MTTR Reduction

0.4%

Supplier Compliance

0.9%

Platform Uptime

0%

Straight-Through

7-14 days<24 hours

CoO Document Cycle

4 hours3 minutes

Cost-Sheet Update

15 min30 seconds

Invoice Generation

25 min10 seconds

Morning Ops Check

FrequentZero

Customs Demurrage

ReactiveSelf-Healing

Incident Resolution

What I'm Looking For

I'm drawn to roles where AI isn't a feature — it's the operating model. Where the challenge isn't "can we add AI?" but "how do we redesign the business around intelligence that compounds over time?"

I bring a specific combination: the ability to understand a P&L, design a system architecture, lead a delivery team, and measure the outcome.

Interested in what I could do for your organization?

Book a 30-Minute Conversation →