Aadil Mansoor
Full profile

Aadil Mohammed Mansoor

Photo of Aadil Mohammed Mansoor

Senior Software Engineer · QBurst Technologies · Kerala, India

Applied-AI engineer and delivery lead who thinks like a solution architect: product first, why before what.

Looking forProject managementProduct ownershipDeveloperArchitect
Email AadilLinkedIn ↗Ask the assistant →
6+

years at QBurst

Since September 2020
6+

engineers mentored

Across projects
3

projects where he approved production releases

Release ownership
69

teams — his team placed first runner-up

QBurst x Alexforbes GenAI Hackathon
Notice period
60 days
Work mode
Prefers remote; willing to relocate for the right offer
Work authorisation
Indian citizen based in India
Compensation
Prefers to discuss directly

What he brings

Applied AI engineering

Agentic AI POCs, MCP-based workflows, tool/function calling, structured outputs, prompt evaluation, grounded report generation, and cost-aware use of AI coding agents.

Technical delivery leadership

Requirements, estimates, task allocation, QA/UAT coordination, production release approval, client communication across time zones, and mentoring at least six engineers.

Backend and data foundation

Python, SQL, REST APIs, B2B integrations, and AWS.

What a recruiter should remember
  • Product first

    He thinks like a solution architect: why something is built matters more to him than what is built.

  • A fresh angle

    He hasn't boxed himself into one mould, so he adapts quickly and looks at problems from different angles.

  • Engineer who delivers

    He has owned estimates, planning, client communication and release approval while still building, and applies AI to real engineering work.

Selected work

Client names stay private (except Alexforbes), so projects are named by what they do.

Supply chainBackend engineer; proactively redesigned the export pipeline

Supply Chain Risk Alert

An alert application that analyses the impact of disasters on facilities, lanes and shipment routes. His clearest example of improving a system nobody asked him to fix.

  • Re-engineered the XLSX/CSV export pipeline with dynamic, reusable queries and a database redesign
  • Found the cause of duplicate rows in batch exports that about 20 developers had missed
  • Tuned RabbitMQ and Celery workers so the client could downsize instances and the queue stopped running out of memory
  • Built Grafana dashboards and alerts for critical shipment and facility events
PythonDjango RESTPostgreSQLPandasRedisCeleryRabbitMQDockerKubernetesAWS RDS/DMSGrafana
2,000+

lines of code removed

Supply-chain export redesign
40%

less module code

Supply-chain export redesign≈ his estimate
25%

better query performance

Supply-chain export redesign≈ his estimate
80%

less effort to add future reports

Supply-chain export redesign≈ his estimate
Before
  • Duplicated per-report export code
  • Duplication errors and missing data
  • Slow exports
After
  • Dynamic, reusable queries
  • New tables without new code
  • Over 2,000 lines removed
Media / salesOne of two developers

Personalised Video Creation for Sales

Automated pipelines that create personalised sales videos, with an on-demand GPU design that only runs when there is work.

  • Cloud Function starts a GPU instance on demand, queues jobs in RabbitMQ, and the instance shuts down after 10 idle minutes
  • Automated website screen recording with Puppeteer and headless browsers
  • Selfie segmentation with ONNX Runtime, frame extraction with OpenCV and Decord, GPU-accelerated encoding
  • Bulk Google Slides personalisation from CSV and Azure text-to-speech
PythonTypeScriptNode.jsPuppeteerFFmpegOpenCVONNX RuntimeRabbitMQGoogle CloudMongoDB AtlasDocker
~80%

less manual editing time

Personalised sales video pipeline≈ his estimate
~40%

lower GPU and cloud cost

On-demand GPU design≈ his estimate
On-demand GPU video pipeline6 steps

“A powerful GPU only costs money while there is work for it.”

  1. 1
    Request
    A personalised video is requested
  2. 2
    Cloud Function
    Takes each request
  3. 3
    Start GPU
    Starts the GCP instance if it is off
  4. 4
    RabbitMQ
    Job is queued
  5. 5
    Poll
    Checked for errors and completion
  6. 6
    Auto-shutdown
    After 10 idle minutes
Financial wellness (Alexforbes)Team lead; owned product design and implementation

AF Engage — GenAI Hackathon

First runner-up among 69 teams at the QBurst x Alexforbes GenAI Hackathon (2026): a B2C financial-wellness prototype for Alexforbes' move from B2B to B2C.

  • Deterministic, UX-led journey under five minutes
  • Calculators, sliders, interactive graphs and progressive disclosure to explain finance without jargon
  • Onboarding, money-personality discovery, guided education, real fund exploration, adviser-led presentations
  • Working prototype in under one week of the two-week hackathon
Claude DesignClaude CodeOpenAI Codex
Advertising / loyaltyDelivery owner and backend developer

Merchandise Verification & Rewards

An app where consumers verify sponsors' merchandise and earn rewards for each verification step.

  • Owned requirements, feature planning, estimation, task allocation and deployments
  • Designed the integration with the client's third-party loyalty-points provider
  • Coordinated frontend engineering and QA; twice-weekly client reviews; led sprint planning
PythonDjangoAWS
~15%

better client-acquisition pipeline

Merchandise Verification & Rewards≈ his estimate

More projects

Pharmaceutical

Pharmaceutical application

Owns user stories end to end on a pharmaceutical application, and helps the team identify agentic-AI use cases and run POCs.

AngularPythonFlaskMongoDBAzure DevOps
Regulated fintech

Money Transfer Application

Full-stack work on a large legacy global money-transfer system, including its migration from Oracle to PostgreSQL.

AngularJSJava 8SpringHibernateOracle DB
Personal finance

Personal Finance Monitoring

Expense tracking with real-time categorisation; he absorbed a late requirements change and still met the original deadline.

Node.jsExpressReactAWS LambdaDynamoDB
Media

Social / Music App

A profanity-free hip-hop music and video app with a social space and video challenges.

Node.jsKnex.jsPubNubAWS S3/EC2/Auto ScalingVimeo
E-commerce

E-commerce Reconciliation

A deterministic reconciliation workflow for high-volume package mapping between Shopify and ShipStation, including gift orders.

PythonShopify APIShipStation API
Business reporting

Business Reporting Pipeline

A web app that compares and aggregates large XLSX files by filters into a new result spreadsheet.

PythonDjango RESTPandas
Advertising

Email Scanner

Extracts the domain from each sender's email address, classifies the website through an API, and lists domains with their classifications.

PythonDjango RESTGoogle APIKlazify
Internal HR

QBTalentHub

Extended an internal talent-management system.

Django REST
Property / booking

Co-working Space Booking

Managed resources and client communication through stakeholder conflicts over budget and scope. More detail is confidential; he can discuss it in an interview.

Personal project

This portfolio assistant

A chatbot that answers questions about him, with layered guardrails so a jailbreak is cheap to fail and gains nothing secret.

OpenUIDeepSeekTypeScriptHono
Personal project (in progress)

Token-efficient code-change routing

Generates an API inventory of a codebase, then uses a small decision-making model to find which microservice needs a change, consulting a larger LLM only where needed.

Skills, honestly rated

Strong = first choice, has started projects in it. Working = productive on real projects. Familiar = reads, debugs and fixes; wouldn't lead a build.

Languages

  • PythonStrong

    Main language and first choice for new work.

  • JavaScript / TypeScriptWorking → Strong

    Across Node.js, Angular and the video pipeline.

  • Java 8 / SpringFamiliar → Working

    About two years; reads and debugs well; last used over 2.5 years ago.

Data

  • SQLWorking → Strong

    Query optimisation on supply-chain exports.

  • PL/SQLFamiliar → Working

    Rusty now; refreshes quickly.

  • PandasStrong

    Around 1 million rows across multiple tables.

  • PostgreSQLStrong

    Data-modelling decisions as solo backend developer.

  • MongoDBWorking

    Current project, small schema additions.

  • DynamoDBWorking

    Rusty; used heavily on one project.

  • RedisWorking

    Part of the supply-chain stack.

  • Polars / PySparkExposure only

    Worked alongside them; never hands-on.

Backend

  • Django RESTStrong

    Started two or three projects from scratch.

  • FlaskWorking

    Works in existing Flask apps; never started one.

  • Node.js / ExpressWorking

    Personal finance, social/music app, Puppeteer automation.

  • REST APIs and B2B integrationStrong

    Rated very strong: led integration architecture on the supply-chain project.

  • RabbitMQ / CeleryWorking

    Designed queuing; tuned memory so the queue never ran out.

Frontend

  • Angular / AngularJSFamiliar → Working

    Own pages in AngularJS; fixes and AI-assisted features in Angular.

  • React / Next.jsFamiliar

    Bug fixes in existing code; not a frontend specialist.

Cloud & ops

  • AWSWorking

    EC2, Lightsail, S3, Lambda, DynamoDB; no Terraform or CloudFormation.

  • Google CloudFamiliar → Working

    GPU Compute Engine, Cloud Functions, Storage.

  • AzureFamiliar

    Cognitive Services TTS; Azure DevOps daily.

  • CI/CDWorking

    Azure DevOps, GitLab, GitHub pipelines; drafts with AI, verifies each.

  • Docker / Kubernetes / HelmFamiliar → Working

    Reads YAML confidently; not an expert.

  • GrafanaFamiliar → Working

    Shipment and facility dashboards plus alerts.

Practice

  • TestingWorking

    pytest; end-to-end API tests with Aloe (BDD); reviews AI-generated unit tests.

  • Video processingWorking

    FFmpeg, OpenCV, Decord; not recent.

Delivery

  • Client handlingStrong

    Weekly and twice-weekly reviews, out-of-hours calls, scope conflicts.

  • Agile / SAFeWorking → Strong

    Sprint and PI planning, estimation, delivery tracking.

How he works

Energised by

Problems that need research, fitting new features into large existing systems, decoding requirements, and client discussions about why they need something.

Presents options, doesn't push

If a client's preferred way is costly, he still offers to build it, but also lays out a cheaper option that meets the need and can be improved later.

Logic over recall

He remembers how a system works, not exact syntax. Once he understands a flow, he pinpoints bugs fast, even in stacks he is less fluent in.

AI-assisted, never blind

He uses AI to draft tests, Dockerfiles and pipelines, then reviews and verifies them himself.

Joins mid-stream

Of his 10+ projects, he joined most midway to fix and improve them; on many he was the solo backend developer.

From requirement to roadmap5 steps

“A mental map of the whole product before anyone estimates.”

  1. 1
    Understand
    The requirement itself
  2. 2
    Write the user flow
    Start to end, every feature and the navigation
  3. 3
    Story spreadsheet
    User-story titles with comments anyone can follow
  4. 4
    Agree
    With the client
  5. 5
    Team estimates
    Story points, plus hours
How he keeps scope under control4 steps

“Everyone knows exactly what is in this sprint, in writing.”

  1. 1
    Define scope
    Exactly what is in this sprint or phase
  2. 2
    Anticipate creep
    Where scope could grow
  3. 3
    Confirm by email
    In writing
  4. 4
    Change request
    Anything new is estimated and billed separately
How he evaluates prompts4 steps

“One judge can be wrong; disagreement is the signal.”

  1. 1
    Three LLM judges
    Score the outputs independently
  2. 2
    Compare
    Look at where they disagree
  3. 3
    Manual review
    Across at least five runs
  4. 4
    Keep the best judge
    The one that matches his judgement for that kind of prompt
Learning right now
  • AI agent guardrails (this chatbot is part of that work)
  • RAG and vector databases
  • Deeper SQL, including verifying AI-generated queries for injection risks
  • A PM or PO certification (planned)

Honest limits

Being clear about gaps is part of how he works.

Not yet: RAG and vector databases

Learning now; not yet built in a project.

Related strength: Claude Certified Architect – Foundations deepened his understanding of how RAG and MCP work behind the scenes.

Not yet: fine-tuning

No fine-tuning or model training yet. He plans to fine-tune Laya, a small decision-making model, for a personal project.

Related strength: He has run pre-trained models for inference, such as ONNX selfie segmentation in a video pipeline.

MCP: daily user, practice builder

Uses MCP servers daily (Azure DevOps, GitHub) but has built them only for practice, not production.

Related strength: He is researching when MCP is worth its token cost versus a ready script or CLI.

Doesn't use Figma

He prototypes with AI design tools instead: Claude Design, Stitch, and his favourite, Open Design.

Related strength: The AF Engage hackathon prototype was a UX-led journey.

Not a syntax memoriser

If a role needs exact syntax from memory, he is not the right fit. His strength is the approach, not recall.

Related strength: Once he understands a flow, he pinpoints bugs fast, even in stacks he is less fluent in.

No PM or PO certification yet

He has done product-owner and project-manager work on most of his client projects, without the formal title or certificate.

Related strength: He wants one and is deciding which would add the most value.

Not a frontend specialist

He handles frontend work, but in React he has done bug fixes in existing code rather than building whole screens.

Related strength: Angular and AngularJS are his stronger frontend tools.

Hasn't owned a budget

He feeds estimates into budgets but has not owned or managed one.

Related strength: He gives effort estimates for budgeting and client-approved scope changes.

Judgement over frameworks

He prioritises with his own judgement and understanding of the project rather than RICE or MoSCoW, and wants to learn the formal methods.

Related strength: He owned backlogs and set priorities on six projects.

No infrastructure-as-code

He designs AWS infrastructure as a developer, not a DevOps specialist, and has not used Terraform or CloudFormation.

Related strength: He has deployed with load balancing and Auto Scaling on EC2 and Lightsail.

Journey

  1. 2016 – 2020
    B.Tech, Electrical and Electronics Engineering
    APJ Abdul Kalam Technological University
  2. Sept 2020
    Joined QBurst Technologies
    Selected through campus placement for his logic skills
  3. 2026
    First runner-up, GenAI Hackathon
    Among 69 teams, QBurst x Alexforbes
  4. July 2026
    Claude Certified Architect – Foundations
    Valid July 2026 to July 2027
Languages
  • EnglishFluent
  • MalayalamFluent
  • TamilConversational
  • HindiBasic speaking, reads and writes
  • ArabicReads and writes, limited understanding
Domains

Pharmaceutical · Supply chain · Fintech / money transfer · Personal finance · E-commerce · Media · Advertising · HR tooling

Get in touch

Notice period 60 days · prefers remote · open to relocating

Got a question? Ask the assistant →