Pijar DSS.
Expansion decisions, quantified.
Summer internship at Pijar, Telkom's edtech arm. I was moved from DevSecOps to the sales team because of my finance background. The head of Pijar handed me the question: when a product isn't selling out, is it always the product team's fault? Or are there other factors? The DSS turns that argument into numbers. 500 Monte Carlo runs over 36 months, VaR, and tornado sensitivity across four scenarios.
Probability, not opinion.
Built after gathering internal data on team capabilities, budget constraints, and market conditions. The engine turns "the product team is the problem" from a hallway argument into a sensitivity chart. Intuition becomes a measurable stance the sales team can bring to the next internal review.
Monte Carlo engine
500+ stochastic runs across a 36-month horizon replace gut-feel go/no-go calls with probability distributions the sales team can actually defend.
Outcome distribution
Profit probability, 2x+ return rate, and ruin risk surfaced side-by-side. Every investment decision answered in probabilities, not vibes.
Return statistics
Mean, median, and win-rate across the full distribution. Lets you separate tails from expectation and see when the headline number lies.
Sensitivity tornado
Pessimistic vs optimistic swings on every input, ranked by impact. Tells you which assumption to pressure-test before signing off on capital.
Risk metrics suite
Value-at-Risk (95%), CVaR, max drawdown, and months underwater. The numbers a finance committee asks for, pre-computed on every run.
Survival curve + presets
Capital-adequacy projection across 36 months, plus Base, Conservative, Aggressive, and Pessimistic presets. Configurable capital, burn, conversion, and PT pricing.
Shipped with the Pijar sales team.
A great and genuinely fun engagement. I was moved from DevSecOps to sales because I cared about finance and wanted to actually help the team. What I got back: a crash course in cross-role collaboration and what working professionally really looks like.
The stack.
Python + FastAPI running the simulation engine. React + Vite + Recharts rendering the cockpit. Fast enough that a 500-run sweep feels interactive.
React
Vite
Recharts
Python
FastAPINeed a decision system like this?
Internal tooling built during an internship at Telkom, so the source stays private. Walkthroughs, architecture notes, and scoped rebuilds are available on request.