Resume
Pranav Sharma
Senior AI Engineer | Forecasting, Pricing Analytics, and Decision Intelligence
Heilbronn, Germany
Email: pranavsharma619@gmail.com · LinkedIn: linkedin.com/in/topranav
I develop deployable AI systems for forecasting, pricing optimization, and industrial decision-making.
My work spans enterprise forecasting algorithms, global sales-pricing optimization, probabilistic ML, constrained decision systems, embedded industrial AI, and AI Operations. I focus on solutions that can be evaluated rigorously, integrated with business rules, deployed reliably, and used in operational decisions.
Selected outcomes include pricing algorithms for approximately 4,000 discount factors across 20 sales organizations, support for approximately 80 hours/month of manual-effort reduction, forecasting models that outperformed existing manual forecasts during validation, contribution to an AI-supported workflow affecting approximately 172 person-days/month of manual effort, and embedded C deployment of fan-grid optimization logic.
Forecasting Sandbox · Pricing Decision Lite · Decision Kernel Lite · Blog · Contact
Profile
Senior AI Engineer focused on operational ML systems for forecasting, pricing analytics, decision intelligence, and MLOps.
Core strengths include:
- demand forecasting systems
- rolling backtesting and model benchmarking
- probabilistic forecasting with quantiles
- conformal evaluation of prediction intervals
- bias-aware evaluation and model governance
- pricing recommendation and optimization workflows
- decision intelligence and risk-aware decision support
- reproducible ML artifacts and deployment support
- AI Operations / AIOps teaching and applied MLOps mentoring
- explaining ML systems to technical and business stakeholders
The common thread in my work is simple:
turn uncertain model outputs into decisions that can be tested, explained, and improved.
Professional Experience
Senior AI Engineer — ebm-papst, Heilbronn
03/2024 – Present
Sales Pricing and Discount-Factor Optimization
- Led the development and validation of pricing algorithms within a global sales-pricing initiative.
- Built ML models to optimize approximately 4,000 discount factors using sales behaviour, actual prices, reference prices, and commercial constraints.
- Automated monthly optimization across 20 sales organizations, replacing a manual process that was difficult to review consistently at scale.
- Built reproducible Kubernetes deployment workflows and supported integration with an operational monitoring dashboard.
- Contributed to standardized pricing logic across sales organizations, supporting approximately 80 hours/month of manual-effort reduction.
Enterprise Forecasting Algorithms
- Led the development and validation of forecasting algorithms within a cross-functional initiative spanning Finance, CRM, and Supply Chain Management.
- Benchmarked statistical, machine-learning, probabilistic, and time-series foundation models using rolling backtesting, demand segmentation, accuracy, bias, and tail-risk analysis.
- Implemented quantile forecasting, conformal prediction, uncertainty quantification, and human-in-the-loop enrichment.
- Built reproducible training and evaluation pipelines; developed models that outperformed existing manual forecasts during validation.
- Contributed to consolidating three manual forecasts into one AI-supported workflow and central dashboard, supporting the transition of approximately 172 person-days/month of manual effort.
- Current development extends the forecasting layer toward feature-based models that integrate commercial and operational business knowledge.
Fan-Grid Efficiency Optimization
- Led the development of an AI-driven optimization solution for scalable fan-grid systems across airflow and pressure operating requirements.
- Designed optimization logic for homogeneous and heterogeneous fan configurations.
- Translated analytical optimization behaviour into a deployable machine-learning model.
- Converted the optimization logic into a lightweight C implementation suitable for embedded ECU environments.
- Defined the problem formulation, operating constraints, validation approach, and deployment architecture.
Business-Constrained Discount Optimization
- Developed a separate constrained-optimization workflow combining data-driven recommendations with explicit business rules.
- Enabled selected discount factors to remain fixed while optimizing only eligible variables within business-defined ranges.
- Integrated configurable commercial values and constraints directly into the optimization logic.
- Preserved expert control by allowing business users to adjust selected parameters and manually refine recommendations.
Lecturer — Hochschule Heilbronn, Heilbronn
AI Operations / AIOps · M.Sc. Business Informatics
2026 – Present
- Teach AI Operations / AIOps for Master’s students in Business Informatics, focused on taking machine learning beyond notebooks.
- Cover practical MLOps foundations including reproducibility, experiment tracking, model packaging, deployment, serving, monitoring, and system reliability.
- Guide students through applied ML workflows: problem framing, dataset selection, baseline modeling, model comparison, inference design, API serving, and deployment readiness.
- Emphasize production-oriented thinking: evaluation discipline, reproducible artifacts, failure handling, documentation, and stakeholder-facing explanation.
- Mentor student teams on building AI systems that can be tested, explained, and presented as operational workflows.
Research Associate — Hahn-Schickard, Villingen-Schwenningen
08/2018 – 02/2025
- Developed deep-learning models for hyperspectral imaging, edge AI, sensor analytics, and resource-constrained ML systems.
- Achieved 91% classification accuracy in hyperspectral skin lesion analysis using transfer learning.
- Contributed to follow-on research funding of more than €1.5M.
- Led development of multi-class deep-learning classifiers for hyperspectral datasets in collaboration with research partners.
- Designed a reinforcement learning agent to improve Kubernetes cluster throughput and operational efficiency.
- Implemented AI methods including clustering, federated learning, explainable AI, and deep learning for imaging and industrial analysis.
- Managed project resources to enable a cost-neutral 6-month project extension despite price increases.
- Supervised and mentored 3 Master’s students.
Assistant Systems Engineer — Tata Consultancy Services, Noida, India
10/2013 – 04/2015
- Supported production maintenance for billing operations software.
- Resolved application issues through structured diagnosis, SQL analysis, and backend investigation.
- Developed frontend components in Qt/C++ and extended backend functionality in SQL.
- Contributed to stable production support in an enterprise application environment.
Selected Portfolio Systems
Forecasting Sandbox Lite
Transparent benchmarking and regime-aware model selection for time-series forecasting.
Focus: rolling backtesting, baseline-vs-ML comparison, MAE and bias-aware scoring, demand regimes, robustness analysis, and audit-ready model selection logic.
Pricing Decision Lite
Scenario-based pricing analytics under uncertainty.
Focus: price-response modeling, guardrails, what-if sensitivity, robust recommendation logic, and decision comparison under uncertainty.
Decision Kernel Lite
Structured decision analysis for comparing actions under uncertainty.
Focus: Expected Loss, Minimax Regret, CVaR, scenario matrices, and risk-aware decision comparison.
Education
M.Sc. Mechatronics Engineering — Hochschule Ravensburg-Weingarten
10/2015 – 03/2018 · Grade: 1.8
- Relevant coursework: Artificial Intelligence, Robot Learning, Robotics
- Master’s thesis: Automatic Maneuver Selection Using Reinforcement Learning, IAV GmbH · Grade: 1.2
- Internship: Automated Markov-based decision-making system at IAV GmbH
B.Tech. Electronics and Instrumentation Engineering — SRM University
07/2009 – 06/2013 · Grade: 7.136
- Relevant coursework: Artificial Intelligence and Expert Systems, Industrial Automation, Engineering Economics and Management
- Internship: Attitude Determination and Control Subsystems, SRM Nanosatellite Lab
Publications
-
Srivastava, A., Sharma, P., Sikora, A., Bittner, A., Dehé, A. Data-driven Modelling of an Indirect Photoacoustic Carbon Dioxide Sensor, IEEE APSCON, 2024. DOI: 10.1109/APSCON60364.2024.10465802
-
Srivastava, A., Sharma, P., Sikora, A., Bittner, A., Dehé, A. Temporal Behavior Analysis for the Impact of Combined Temperature and Humidity Variations on a Photoacoustic CO₂ Sensor, IEEE APSCON, 2024. DOI: 10.1109/APSCON60364.2024.10465885
-
Sharma, P., Rüb, M., Gaida, D., Lutz, H., Sikora, A. Deep Learning in Resource and Data Constrained Edge Computing Systems, Machine Learning for Cyber Physical Systems. DOI: 10.1007/978-3-662-62746-4_5
Awards and Honors
- Certificate of Commendation — SRM University and Indian Space Research Organization
- Founder’s Scholar — SRM University
Certifications
PMP · IBM Generative AI Engineering with LLMs · DeepLearning.AI Deep Learning Specialization · Udacity Computer Vision Expert Nanodegree · Google Advanced Data Analytics · Duke AI Product Management · DeepLearning.AI GANs
Skills
Core Domains
Demand forecasting · Pricing analytics · Decision intelligence · Probabilistic ML · AI Operations / MLOps · Model evaluation and governance
Forecasting and Uncertainty
Statistical baselines · ML forecasting · Quantile forecasting · Conformal prediction · Rolling-origin backtesting · Bias-aware evaluation · Model robustness analysis · Demand regime segmentation
Decision and Pricing Systems
Scenario analysis · Sensitivity analysis · Expected Loss · Minimax Regret · CVaR · Guardrailed recommendation logic · What-if analysis
Technical Stack
Python · SQL · pandas · NumPy · SciPy · scikit-learn · statsmodels · LightGBM · XGBoost · PyTorch · Docker · Kubernetes · MLflow · DVC · FastAPI · Streamlit · Hugging Face Spaces · Git/GitHub · Linux/CLI
Contact
For roles, technical collaboration, consulting discussions, or advisory conversations:
Email: pranavsharma619@gmail.com
LinkedIn: linkedin.com/in/topranav
Compliance Note
This website contains independent portfolio and profile content only. Views are my own and do not represent any employer. No proprietary or confidential information is included.