Saqib Ameen Subhan, Database Engineering Leader

01Measurable Impact

Numbers that matter

Infrastructure Cost Reduction

Led the migration of one monolithic MongoDB cluster (96 cores, 768GB RAM) into 65 right-sized clusters segmented by customer tier. Zero customer downtime across 8 months. $350,000 saved annually. Recognised with the company Thank You Award.

0hrs

Regional Disaster Recovery

Full database estate recovery after AWS ME-Central-1 went completely offline during the Middle East conflict. Navigated UAE Central Bank data-residency constraints under emergency conditions. Database restore in 5 hours; back online before the banks were.

0min

Autoscaling Pipeline

End-to-end automated database scaling: CloudWatch → SNS → SQS → rolling node upgrades with health-check validation at every step. Zero application impact. The on-call engineer became a notifier, not an actor.

Nodes Managed at Visa

Subject-matter expert for 1,000+ node Cassandra and MongoDB environments across multi-data-centre, multi-region private cloud, at global financial transaction scale.

Self-Serve

Database Provisioning Portal

Built a self-service portal at Visa: developers request environments through a UI, automation provisions them, and TTL auto-decommission cleans them up. The DBA bottleneck was removed from the developer workflow entirely.

0×

Certified Database Expert

MongoDB Certified DBA · MongoDB Certified Atlas Administrator · MongoDB Certified Associate Data Modeler · Apache Cassandra 3.x Administrator (DataStax).

02Open Source

mdbkit, the tool I got
annoyed enough to build

mtools was brilliant, and its log tools never learned to read the JSON log format MongoDB introduced in 4.4. So I created the replacement.

mdbkit is an offline toolkit for MongoDB structured logs: slow-query analysis, deterministic index advice, incident triage and FTDC decoding, from the terminal, for MongoDB 4.4 through 8.0.

The part I care most about: it never connects to your database. No driver, no URI, no network code anywhere in it. Zero runtime dependencies, nothing in the supply chain but the Python standard library. It reads files, it is strictly read-only, and where an action would help it prints the command for you to review and run yourself. The README shows you how to verify all of that in a minute with grep and strace.

And when it recommends an index, that advice is rules, not AI. The same log always produces the same recommendation, with the evidence it reasoned from, a confidence level, and how to validate it. It says candidate, never command, and it will never tell you to drop an index.

PyPI MIT Licensed Zero Dependencies Offline by Design MongoDB 4.4 to 8.0 Python 3.8+

# try it in sixty seconds, without touching a cluster $ pip install mdbkit $ mdbkit demo -o demo.log $ mdbkit queries demo.log namespace op count cumMs scan plan shop.events aggregate 29 3.2m 98889:1 COLLSCAN+SORT shop.orders find 48 1.4m 2976:1 COLLSCAN+SORT shop.users find 31 3.7s 1:1 IXSCAN{email} # two of those need an index. one is already fine. # that distinction is the whole point.

GitHub PyPI

Why is it slow?$ mdbkit queries mongod.log --sort scanRatio $ mdbkit advise mongod.log --indexes indexes.json

Slow queries grouped by query shape, ranked by total time and examined-to-returned ratio/deterministic ESR index candidates with evidence, confidence and a validation step

What happened at 3am?$ mdbkit triage mongod.log --oslog /var/log/syslog $ mdbkit triage mongod.log --report incident.html

Elections, connection storms, index builds, error clusters and slow-query peaks in one screen/correlates the OOM kill in the system log with the restart mongod could never record itself

Did my change help?$ mdbkit compare before.log --after after.log $ mdbkit explain explain.json

Diffs query shapes between two logs: improved, regressed, new, gone/reads a saved executionStats plan and does the examined-versus-returned maths for you

What do the metrics say?$ mdbkit ftdc summary diagnostic.data --last 4h $ mdbkit serverstatus before.json --after after.json

Decodes diagnostic.data offline: CPU, cache, queues, tickets, connections, with no agent installed/digests serverStatus down to the handful of fields that explain a struggling server

Run it anywhere$ mdbkit demo --scenario incident -o demo.log $ mdbkit lab start && mdbkit lab seed $ mdbkit triage mongod.log --only CRIT,WARN --exit-code

Deterministic synthetic incident logs for evaluation and demos/a disposable local replica set for before-and-after experiments/exit codes that make it a cron job, and self-contained HTML reports that open air-gapped

03AI & Agents

I don't just use AI.
I build with it

Fourteen years of database engineering taught me how systems behave under real load. I apply the same rigour to AI, building agents that solve my own problems, and bringing AI-native workflows into how I engineer, architect, and operate data platforms.

001

Vector Search at Scale● Production Architecture

Shipped MongoDB Atlas Vector Search to production at Contentstack for an AI-driven content generation feature, evaluating Pinecone and Milvus before choosing Atlas for retrieval performance and data co-location. Real RAG architecture serving real users, not a demo.

002

Self-Hosted Agent Infrastructure● Autonomous · Running

Agents running on my own VPS, model-routed and scheduled, doing continuous research and synthesis work that would otherwise be a daily manual grind. Built end to end: provisioning, model routing, cost control, and failure handling, because an agent that cannot recover from a bad response is a demo, not a system.

003

Health Agent● Personal Agent · Production

A personal health agent that reasons over years of my own medical history and gives real-time dietary guidance. Send it a photo of a meal, get an instant call on whether it fits my profile. Deliberately self-hosted on my own infrastructure, keeping sensitive data under my own control rather than handing it to a third party.

The Argument

Your AI is describing a MongoDB that no longer exists. It is not lying to you. It is describing roughly 4.2, because that is where the internet's supply of MongoDB blog posts is centred. Learn to read the plan, and the model stops being a confident guesser and becomes a very fast reader, which is what it is actually good at.

From the Dubai MongoDB User Group talk, August 2026. Cardinality, write mix, and what is in your plan cache right now are three things a model can never know from your query text, and all three are printed, for free, in the output of one command.

04whoami

Based in Bengaluru, India
Previously Abu Dhabi & Singapore

Built on depth, not just breadth. Fourteen years at the intersection where databases meet business outcomes.

From classified government networks in Abu Dhabi to FinTech platforms processing millions of transactions, from 1,000-node Cassandra clusters at Visa to 65-cluster migrations at high-growth SaaS, I have operated in environments where database failures are not an option. Today I lead data services engineering at IG Group, a global FinTech with a fifty-year history and operations across eighteen countries.

My approach is simple: understand the business problem first, then select the right technology. I have worked across MongoDB, Cassandra, Oracle, PostgreSQL, ClickHouse, Aurora, MySQL, Redis, Neo4j, and SQL Server, self-managed and cloud-native. Each has its place. The skill is knowing which place.

I also own the cost and vendor decisions that come with the platform. I have run multi-year TCO analyses: negotiating tooling contracts, weighing managed services against self-managed infrastructure, and making the call that saved real money without compromising reliability. Platform ownership is as much a commercial discipline as a technical one.

And I write and ship in the open. mdbkit is on PyPI under MIT, Tales from Production runs on dev.to, and I co-lead Namma MUG in Bengaluru after founding the MongoDB Abu Dhabi User Group. None of it depends on where I work.

Across sectors: FinTech · SaaS · Banking · Trading · Gaming · Government · Social Media · Payments

MongoDB Community Creator 2024, 1 of 45 globallyMongoDB Inc

Creator of mdbkit, open-source MongoDB log toolkitMIT · PyPI

MongoDB Certified DBA v4.2MongoDB Inc

MongoDB Certified Atlas AdministratorMongoDB Inc

MongoDB Certified Associate Data ModelerMongoDB Inc

Apache Cassandra 3.x Administrator AssociateDataStax

LinkedIn Top Database VoiceLinkedIn

Co-Leader, Namma MUG (MongoDB User Group), BengaluruCommunity

Founding Leader, MongoDB Abu Dhabi User GroupCommunity

05Technical Depth

What I build and operate

Distributed Data SystemsCore Databases

MongoDB Atlas/Cassandra/Oracle/PostgreSQL · Patroni/Aurora PostgreSQL & MySQL/ClickHouse/Redis/Neo4j/MS SQL Server

Cloud & On-PremInfrastructure

AWS (deep): EC2 · S3 · IAM · CloudWatch · SNS/SQS · DMS/On-prem & private cloud/Nutanix HCI/Hybrid architecture/GCP & Azure

Automation & DevOpsPlatform Engineering

Python / boto3/Shell scripting/Terraform/CI/CD/PGBouncer/Debezium · CDC/Apache Kafka/TCO optimisation/DR · RTO/RPO design

AI-Native OperationsAI & Emerging

Atlas Vector Search/Pinecone · Milvus/RAG architectures/Agentic systems/LLM integration/Claude · Codex · Gemini

Engineering LeadershipPeople & Platform

Building teams from scratch/Mentoring and promoting engineers/24×7 on-call and runbook culture/Cross-functional alignment/Roadmap planning/Budget and COGS ownership/Regulatory and audit engagement

06Career Journey

14 years of consequential work

Aug 2026 to PresentBengaluru, India

IG Group

Engineering Manager, Data & AI

Leading a high-performance, global data services team transforming enterprise-scale data infrastructure at a fifty-year-old FinTech operating across 18 countries
Driving modernisation, automation, and strategic cost reduction across the database estate

Mar 2025 to Aug 2026Abu Dhabi, UAE

Avrioc Technologies LLC

Senior Database Architect

Architected database platforms across 4 concurrent product lines: FinTech (UAE Central Bank regulated), Social Media, Virtual Cycling/Gaming, and Communications
Led full database estate recovery in 30 hours when the AWS ME-Central-1 region went completely offline during the Middle East conflict
Navigated UAE Central Bank data residency regulations under emergency conditions to obtain cross-region restore approval
Migrated high-traffic social media product from AWS RDS MySQL → Aurora Provisioned → Aurora Serverless with zero downtime, enabling sub-second autoscaling for unpredictable celebrity-driven traffic spikes
Drove ClickHouse adoption from PoC to production for real-time analytics: 10× data compression, sub-millisecond queries; expanded into the FinTech payments product
Designed and operated self-managed PostgreSQL HA clusters on Patroni: three nodes per data centre across two DCs with F5 load balancing, alongside managed Aurora PostgreSQL for a cost-optimised hybrid footprint
Built and managed a multi-product database team delivering 24×7 operational support across Aurora, RDS, ClickHouse, self-managed PostgreSQL, Cassandra, MongoDB, and Redis

Jan 2022 to Feb 2025Bengaluru, India

Contentstack Inc.

Senior Manager, Databases · Engineering Manager, Databases

$350K saved annually. Led an 8-month migration from a single monolithic MongoDB cluster (R5.24xlarge, 96 cores, 768GB RAM) to 65 right-sized clusters segmented by customer tier, with zero customer downtime and a 48% reduction in database infrastructure COGS
Built in-house MongoDB autoscaling pipeline using CloudWatch, SNS, SQS, and AWS CLI, with rolling node upgrades in 12 minutes with health-check validation at each step
Led MongoDB Atlas Vector Search to production for AI-driven content generation: evaluated Pinecone and Milvus, chose Atlas for retrieval performance and data co-location
Ran the Cloud Manager vs Atlas TCO analysis with MongoDB Professional Services, the multi-year infrastructure decision the company committed to
Owned database platforms serving ~2 billion API calls a day across a global multi-tenant SaaS estate
Built the database team from scratch: two engineers hired as interns now lead teams of their own
Ran weekly Tech Byte sessions with product engineering: schema-related production incidents dropped ~70% in one quarter

Jul 2019 to Jan 2022Singapore & Bengaluru

Visa Inc.

Associate Manager / SME, NoSQL Databases

SME for 1,000+ node Cassandra and MongoDB environments across multi-data-centre, multi-region private cloud at global financial transaction scale
Built a self-serve database provisioning portal: request via UI, automated provisioning, TTL auto-decommission; eliminated the DBA bottleneck entirely
Deployed MongoDB CloudManager/OpsManager with enterprise security: SSL/TLS, KMIP, LDAP
Multiple GoBeyond Awards for large-scale NoSQL platform engineering

May 2018 to Jun 2019Bengaluru, India

Accenture Services

NoSQL Databases Team Lead

Led a 5-member team supporting a premier US financial services client across MS SQL, MongoDB, and Cassandra
Delivered the NoSQL transition for a UAE government entity: POCs, stakeholder trust building, transition approval secured. Core Value Champion Award

Dec 2016 to Apr 2018Abu Dhabi / Dubai, UAE

Emaratech

Database Administrator

Mission-critical government databases across Oracle, MSSQL, and Cassandra, with 24×7 support under stringent SLAs
GEM Award for a novel Oracle GoldenGate monitoring mechanism over DataDiode in classified government networks

Aug 2012 to Nov 2016Bengaluru, India

Societe Generale

Database Administrator

Business-critical production databases across Oracle, Cassandra, SQL Server, and Sybase, with 24×7 global banking support
Data Guard and SQL Server Mirroring for DR; Oracle GoldenGate for cross-platform replication

07Recognition

Work that gets noticed

MongoDB Community Creators Worldwide

The MongoDB Community Creator program selects fewer than 50 engineers globally each year. I am one of them, among the very few from India and the UAE. There was a MongoDB user group in Dubai and nothing in Abu Dhabi, so I was asked to start one. I founded it and ran its first session as its founding leader. I now co-lead Namma MUG in Bengaluru, speak at community and industry events, and publish regularly on database architecture, performance, and AI-native data systems. Community building is not something I do for recognition. It is how I stay sharp.

MongoDB Community CreatorMongoDB Inc · 1 of 45 globally, Champion Finalist2024

Founding Leader, MongoDB Abu Dhabi User GroupStarted the group where none existed and ran its first session2025

Co-Leader, Namma MUG BengaluruNamma MUG (MongoDB User Group), Bengaluru2026

Thank You AwardContentstack · 48% COGS reduction, $350K annually2023

GoBeyond Award × MultipleVisa · large-scale NoSQL platform engineering2020-21

Core Value ChampionAccenture · US financial services client2019

GEM AwardEmaratech · GoldenGate over DataDiode innovation2018

Top Database VoiceLinkedIn · community designationOngoing

Saqib Ameen Subhan