TheFifthPostulate

Applied Decision Science Portfolio

Uncertainty-Aware Modeling and Human-in-the-Loop Systems

Overview

I am an applied decision scientist building uncertainty-aware forecasting and decision support systems for high-stakes, data-scarce environments. I specialize in Bayesian modeling, inventory analytics, and human-in-the-loop workflows across regulated domains such as biotech and healthcare.

My work emphasizes principled modeling, honest failure analysis, and designing analytical systems that make automated inference workflows reliable under uncertainity.

Design Philosophy

Most data science portfolios optimize for predictive accuracy in static benchmark settings. My work instead focuses on decision-grade reliability in real operational environments characterized by sparse data, nonstationarity, and asymmetric failure costs.

Across projects, I design systems that:

• Explicitly model uncertainty rather than collapsing it into point predictions
• Encode operational constraints into decision rules
• Intentionally throttle automation when model confidence is insufficient
• Integrate human oversight as a first-class component of the system
• Treat model failure as a diagnostic signal about system structure, not a defect to be hidden

This perspective is informed by experience in regulated domains where false confidence, silent failure, and automation bias carry real-world consequences.

Project 1 — Decision-Aware Stochastic Consumption Forecasting

This project explores stochastic inventory forecasting under severe covariate scarcity using Poisson–Gamma conjugacy and a waste-constrained restocking policy.

It demonstrates both a principled Bayesian modeling approach and the structural limits of automated forecasting in nonstationary, human-driven consumption systems.

Key contributions:

• Exposure-normalized Poisson–Gamma modeling of consumption rates
• Closed-form posterior predictive forecasting via Negative Binomial distributions
• Monte Carlo decision policy optimizing reorder quantities under waste constraints
• Rolling-origin evaluation under regime shifts
• Formal falsification of automation viability under nonstationarity

Artifacts:

• 📄 Methods Note (PDF): https://thefifthpostulate.github.io/projects/stochastic-forecasting.html
• 📊 Analysis Notebook: https://thefifthpostulate.github.io/Stochastic-Consumption-Forecasting/InventoryProject.html
• 💻 Source Code: Available upon request

Key takeaway:
Uncertainty modeling revealed the true complexity of the consumption process. Assumptions about the stochastic process and decision rule were insufficient to consistently provide decision-grade forecasts, making expert oversight more reliable than fully automated inventory control.

Project 2 — Geometry-Based Risk Modeling for Safe ML Inference in Medical Diagnostic Triage

This project develops a safety-aware diagnostic inference system on the Breast Cancer Wisconsin dataset that explicitly models and intercepts failure modes in probabilistic classifiers. Rather than optimizing for overall accuracy, the system introduces a post-classification reliability layer that detects high-risk predictions using geometry-derived signals and routes them to human review, enforcing a zero–false-negative constraint in clinically ambiguous regions. The core contribution is a selective inference control system that extracts additional risk signals from the information geometry of class-conditional feature manifolds, enabling reliable automation in the presence of deep class overlap and overconfident model failures.

Key contributions:

• Geometry-derived risk signals that expose hidden false-negative failure modes beyond model probabilities
• Selective inference policy that enforces zero false negatives via principled abstention
• Reliability layer decoupled from the base classifier, enabling controllable human-in-the-loop deployment
• Bootstrap-validated decision thresholds for stability under retraining and sampling uncertainty
• Quantitative analysis of automation–review tradeoffs in safety-critical inference

Artifacts:

• 📄 Methods Note (PDF): (link)
• 📊 Analysis Notebook / Demo: https://thefifthpostulate.github.io/Geometric-Risk-Modeling/geometric_risk_modeling.html
• 💻 Source Code: Available upon request

Key takeaway: By explicitly modeling geometric failure modes and enforcing selective abstention, this system transforms a high-performance classifier into a controllable decision system with principled human oversight, illustrating a general framework for deploying machine learning safely in high-stakes environments.

Contact

Jithakrishna Prakash
📧 jprakashoff@gmail.com
🔗 LinkedIn: https://linkedin.com/in/jithakrishna-prakash
💻 GitHub: https://github.com/TheFifthPostulate