Monu Kumar Pathak

Hi, I'm Monu Kumar Pathak

Applied AI and Algorithms Engineer.

Currently working as an SDE 1 (Applied AI & Algorithms) in the Kinaxis India Research Team.

I completed my M.Tech in IEOR at IIT Bombay, specializing in computation and data science. My work spans reinforcement learning, agentic decision intelligence, simulation, and stochastic optimization.

My usual day looks like this:
• Teaching machines to think.
• Occasionally teaching them not to hallucinate.
• Constantly teaching myself to do the same.

Monu Kumar Pathak - AI and LLM Researcher
0
Projects Completed
0
Technologies
0
Research Areas

Skills & Technologies

🤖 AI & Machine Learning

PyTorch TensorFlow Scikit-learn Deep RL SHAP

📝 NLP & LLMs

LangGraph Transformers GPT Grounded Generation Human-in-the-loop AI

💻 Programming

Python SQL Bash Pandas NumPy

🔧 Tools & Platforms

Gurobi SciPy CPLEX SimPy AnyLogic

Career Trajectory

Kinaxis India

Current phase · Jul '25 – Present

FDE, Agentic Solutions

Algorithms Engineer

MEIO (Optimization and Simulations)

My responsibilities include:

  • Developing an agentic supply-chain decision intelligence platform
  • Developing customer-support agents that explain inventory optimization recommendations
  • Adapting foundation models for time-series forecasting
  • Applying reinforcement learning to inventory optimization
  • Developing solver modules for inventory optimization
  • Performing statistical distribution modeling and goodness-of-fit validation

FJSSP Simulation & RL for Production Scheduling - IIT Bombay

Research phase · Jan '25 – Present

Built an open-source SimPy library for flexible job-shop scheduling and a DQN agent that reduced makespan by 42% versus conventional dispatching rules.

Research topics:

  • Open-Source FJSSP Discrete-Event Simulation Library
  • Reinforcement Learning for Production Scheduling Systems

Master's Thesis - XAI in Simulation Output Analysis

Research phase · Jun '24 – Jan '25

Ran more than one million simulations and combined data farming, clustering, neural networks, random forests, and SHAP to explain simulation outcomes.

EHR-based Patient Risk Modeling - Cognitio Analytics

Internship phase · May '24 – Jul '24

Modeled 58,000 hospital admissions with LSTM, Transformers, and LightGBM, improving disease-classification AUROC by 19% over end-to-end neural models.

Selected Projects

Agentic Supply-Chain Decision Intelligence

Built a grounded, auditable LangGraph decision layer for supply-chain data and live events.

Developed a stateful, checkpointed agentic system with human approval for every data write. Intent-driven selective retrieval achieved a 93% evaluation pass rate, 96% grounded structured output, 2.4-second median latency, and reduced numeric hallucination from 18% to 2%.

Sample-Efficient Inventory Policy Learning

Learned multi-echelon replenishment policies through differentiable simulation.

Built a PyTorch inventory model using HDPO and implemented vanilla and graph neural networks to coordinate store and warehouse replenishment. Achieved optimality gaps as low as 0.03% while reducing data requirements by up to 64x.

Deep RL for Production Scheduling

Reduced flexible job-shop makespan by 42% with a custom DQN scheduling agent.

Modeled a production scheduling system as a discrete-event simulation in SimPy, implemented a DQN from scratch, and trained it with custom reward shaping to outperform conventional dispatching rules.

XAI for Simulation Output Analysis

Explained complex simulation outcomes using data farming, clustering, and SHAP.

Performed more than one million simulations, clustered outputs with K-Means and DBSCAN, and used SHAP to quantify each input feature's marginal contribution across neural-network and random-forest models.

Differentially Private Recommendation Systems

Built an open-source recommender with privacy protection across the full pipeline.

Applied three Laplace-noise mechanisms at input, processing, and output stages and implemented ALS and SGD matrix factorization from scratch. Achieved 0.92 RMSE on perturbed data versus 0.85 on clean data.

Commonsense Reasoning in LLMs

Fine-tuned GPT-1 on large knowledge graphs for commonsense inference generation.

Leveraged ConceptNet and ATOMIC knowledge graphs with over 40 million and 877,000 tokens respectively, achieving 11.14 perplexity and a 15.10 BLEU-2 score for generated commonsense inferences.

Multi-Echelon Inventory Network Modeling

Modeled stochastic demand and lead-time propagation across complex inventory networks.

Derived lead-time demand moments under demand and lead-time uncertainty, selected suitable probability distributions, and computed service-level-constrained reorder points and expected backorders. Modeled bill-of-material dependencies as a graph with ranking and cycle-detection algorithms.

Derivative-Free Inventory Optimization

Optimized service levels in nonlinear and discontinuous inventory systems.

Formulated a stochastic nonlinear cost model and solved it with SciPy's Dual Annealing and Differential Evolution. Added batched warm starts and cost-based variable elimination to accelerate convergence and improve numerical stability.

Statistical Distribution Modeling

Built and validated parametric approximations for convoluted probability distributions.

Developed a parametric approximation for the convolution of independent distributions, ranked candidate models using AIC and BIC, evaluated fit with the Kolmogorov-Smirnov statistic, and validated results with PP and QQ plots.

Open-Source FJSSP Simulation Library

Created a configurable SimPy library for flexible job-shop production scheduling.

Supports configurable work centers and machines, pluggable FIFO, SPT, EDD, LPT, and FIS sequencing rules, and complete live-state access. Simulations can be paused, tested through sub-simulations, and resumed with the selected strategy.

EHR-Based Patient Risk Modeling

Predicted patient outcomes from 58,000 hospital admissions and longitudinal medical histories.

Preprocessed EHR data for more than 38,000 adult patients, developed LSTM and Transformer models for next-visit embeddings, and classified diseases into five groups. LightGBM improved AUROC by 19% over end-to-end neural models.

Insurance Risk Quantification

Modeled insurance claim distributions and quantified tail risk with VaR and TVaR.

Analyzed claim amounts using statistical visualizations, identified underlying distributions with Kolmogorov-Smirnov and chi-squared tests, compared risk across claim categories, and proposed no-claim incentives and tiered premium pricing.

Endorsements and Recommendations

"Monu possesses deep knowledge of LLM and simulation research. He helped me many times during college and continued to provide that expertise in the industry."
ViplawOnSemi Research
"Monu brings a strong statistical foundation to every project. While working together on distribution modeling, his ability to break down mathematical complexities into functional logic made him a highly effective collaborator."
SaurabhData Scientist - IndusInd Bank

Get In Touch

Open to collaborations and interesting challenges.