Team
Fifteen people: three managers, four senior analysts, and analysts embedded in each business line rather than pooled behind a central queue.
Leading Business IntelligenceBI at FundedNext
I built FundedNext's data function from zero.
I'm Kazi Nadimul Haque, Manager II of Business Intelligence at Next Ventures. I joined in 2020 as the first analyst, into a company with no warehouse, no pipelines, and no reporting. Today I lead fifteen people across CFDs, Futures, and Broker, and own product for CFDs.
LeadManager II
1
ManagersManager I
3
Senior analystsSr. Executive
4
AnalystsExecutive
8
20201 analyst, no data function
Today15-person team, 3 business lines
- 15Team members across two management layers
- 3Business lines: CFDs, Futures, Broker
- 10M+Rows processed every day
- 100+Dashboards in production
- 7+Years across analytics and commercial roles
01 About
One analyst to a fifteen-person function, through hypergrowth.
When I joined Next Ventures in 2020, there was no warehouse, no pipelines, and no reporting layer. Numbers came from whoever could export them. I was the data function.
In the years since, FundedNext has grown into a global prop trading firm, and the data function grew with it — from one analyst to fifteen people across two management layers, with analysts embedded directly in the CFDs, Futures, and Broker business lines.
My job moved from building dashboards to building the team, standards, and platform that let other people do it well. I also took on product ownership of CFDs, so I'm accountable for the decisions the data informs — not just the data.
Before data, I worked in sales at Grameenphone and Toshiba. That's where the commercial instinct behind my analytical work comes from.
02 Scope
What I own today
Four areas where the decision stops with me.
Domains
Data, insight, and decision support across the CFDs, Futures, and Broker business lines.
Product
Product Owner for CFDs. Rule changes and structural modifications to the product run through me.
Platform
A PostgreSQL warehouse fed by Airflow-orchestrated ELT, processing tens of millions of rows a day into 100+ dashboards.
03 Platform
From raw data to decisions
The stack my team builds and runs, from operational sources through to the three business lines that act on it.
01 · Sources
Operational data
- Product & trading data
- Payments
- KYC & onboarding
02 · Orchestrate
Apache Airflow
- Scheduled ELT pipelines
- SQL transformations
03 · Store
PostgreSQL warehouse
- 10M+ rows a day
- Semantic data models
- Explicitly named metrics
04 · Serve
BI layer
- Power BI
- Apache Superset
- Metabase
- 100+ dashboards
05 · Decide
Business lines
- CFDs
- Futures
- Broker
- Risk & Compliance
Alongside
ML anomaly detection
A/B testing
LLM-assisted reporting
04 Work
Selected work
The problems my team and I have spent the most time on.
Case study · 7 min read
Read the case study
Building a BI function from zero
How the data function grew from one analyst to fifteen people across three business lines — the warehouse call, the pass-rate definitions, the model nobody used, and what I would do differently.
Product analytics
Trader lifecycle analytics
End-to-end analytics across acquisition, onboarding, evaluation, and funded stages — the basis for decisions on activation and evaluation-to-funded conversion.
Named metrics, not one number
Three business lines counted pass rate three ways. Instead of forcing a single number, we kept each variant and named it explicitly, so a label always means one calculation.
A/B tests on pricing and rules
Designed and evaluated experiments on pricing, challenge rule structure, and feature releases, and turned the results into rollout decisions.
ML anomaly detection
Anomaly detection and rule monitoring built with the Risk and Compliance teams to streamline monitoring and reporting.
LLM-assisted reporting
Applying large language models to internal reporting workflows, moving recurring write-ups off analysts' plates.
05 Principles
How I work
What building a data function from zero, during hypergrowth, taught me.
Metric definitions are a governance problem, not a technical one.
The hardest thing I have enforced is not a pipeline or a schema. It is getting three business lines to stop using one name for three different numbers. The SQL was the easy half.
Embedded beats centralised, once you can afford it.
A central queue optimises for analyst utilisation. Embedding optimises for decisions actually changing. Those are different goals, and the second is the one that matters.
Adoption is the deliverable.
I have shipped a model that was technically sound and went unused, because the question it answered had moved on by the time it landed. Check the question is still live before you finish, not just before you start.
Demand will always exceed capacity.
The answer is not heroics. It is a legible intake process, an honest backlog, and the willingness to tell stakeholders what their request costs and what it displaces.
06 Experience
From the sales floor to data leadership
-
2020 — Present
Analyst → Manager II, Business Intelligence
Next Ventures (FundedNext)
Built the BI function from zero. Lead a fifteen-person team across CFDs, Futures, and Broker, and serve as Product Owner for CFDs.
-
Dec 2019 — Feb 2020
Senior Executive, Corporate Sales
International Office Machine (Toshiba)
B2B account management, prospect evaluation, and sales reporting.
-
Jul 2016 — Feb 2019
Senior Trainee, Sales & Operations
Grameenphone
Customer relationships, negotiation, and market feedback at Bangladesh's largest mobile operator.
Employee of the Month ×9GP 4G National ChampionshipBattle of Excellence · Best PerformerGP Intra 4G Championship
Education
- MSc, Computer Science & EngineeringData Science · United International University2024
- BBA, FinanceEast West University2019
Toolkit
SQLPostgreSQLMySQLPythonRAirflowPower BISupersetMetabaseGCPMachine learningLLMsGit
07 Contact
Let's talk data leadership.
Open to conversations about building and scaling data functions in fintech and trading.
linkedin.com/in/kazi-nadimul-haque · github.com/nadim-raj · nadim-raj.github.io