Samarth Mamadapur

Machine Learning Engineer

um.samarth@gmail.com+49 17686659027LinkedIn
Amsterdam, Netherlands
Senior ML Engineer on the Applied AI team at Uber, where I build intelligent agents that empower Uber Eats merchants. With 6+ years of experience shipping production ML/AI systems, I specialize in translating ambiguous product problems into reliable, data-driven solutions. Previously drove data science initiatives for the micromobility vertical atBolt and partnered with leading pharmaceutical companies while at ZS.

About Me

Passionate about productionizing AI systems, from computer vision to LLMs. I also enjoy communicating complex results to both technical and non-technical audiences while driving product and business impact.

Experience

Uber
Senior ML Engineer, Applied AI
August 2025 – Present
Amsterdam
Bolt
Data Scientist
June 2022 – July 2025
Berlin
  • One of the earliest Data Science hires for the Micromobility vertical at a fast-growing European mobility startup.
  • Scaled DS team and initiatives while serving millions of users across 100+ cities.

Pricing Optimization Strategy

  • Architected a multi-model (including Bayesian hierarchical) forecasting system with business-knowledge-based priors and constraints to optimize pricing across pay-as-you-go rides and subscription passes.
  • Achieved a significant MAPE reduction from 14.14% to 8.7% while improving causal forecasting quality.
  • Built robust forecasting models using a Bayesian framework, handling multiple pricing scenarios across 100+ markets, guiding vertical-wide revenue decisions.
  • Reduced average forecast MAPE by ~5% across all cities, leading to much higher-quality forecasts and downstream pricing decisions based on the model.
  • Translated an extremely ambiguous business challenge (pass vs pay-as-you-go) into a structured, data-driven solution.
  • Designed an experimental framework to measure price elasticities and cross-elasticities between different product offerings.

Smart Collect

  • Led end-to-end development optimizing operational efficiency.
  • Defined the north-star metric (GMV per Collection Task) through stakeholder alignment.
  • Designed rigorous A/B testing with user-centric guardrails.
  • Drove a 2.7% improvement in the primary metric while maintaining user satisfaction thresholds.
  • Orchestrated the ML-based solution to a complex operations process end-to-end.

Computer Vision Solutions

  • Led ParkAssist project: defined parking quality standards, coordinated labeling efforts, and fine-tuned vision models to achieve 85%+ accuracy.
  • Processed millions of weekly parking validations across 100+ cities, improving operational efficiency and user experience.
  • Achieved best-in-class parking detection models (across all scooter competitors, from public tender evaluations).
  • Allowed operation in new markets and helped win tenders in existing markets, indirectly resulting in millions of new rides.
  • Developed tandem riding detection system using a first principles approach, crucial for regulatory compliance and winning city tenders.
  • Established a data-driven product development culture by operationalizing vertical-wide standard metrics for experiments and holding weekly data insights sessions with product and engineering teams to inform decision-making.
December 2023 – Present
Remote
  • Built and launched a production AI application that generates album artwork by chaining LLMs with diffusion models.
  • Designed and implemented end-to-end system architecture, from prompt engineering to model deployment.
  • Developed full-stack infrastructure using NextJS, handling user authentication and payment processing.
  • Engineered robust prompt optimization and model chaining pipelines for consistent, high-quality outputs.
  • Demonstrated ability to take an AI project from concept to monetized product, gaining hands-on experience with LLM development and deployment.
Carnot Technologies
Data Scientist
October 2021 – April 2022
Remote
  • Sole Data Scientist working on agri-tech solutions for a Mahindra-backed startup.
  • Wrote production-level Python code to analyze large-scale, noisy IoT data from on-board devices (OBDs) on tractors (2Hz sampling rate) to detect fuel theft.
  • Performed product usage analytics for an app with over 1 million registered users.
View Earlier Experience

Skills & Tools

Technical Skills

Machine Learning
LLM Engineering
Deep Learning
Computer Vision
Statistical Inference
Experiment Design / A/B Testing
Model Deployment
Full Stack Development

Tools & Technologies

Python
SQL
NextJS
AWS
Docker
OpenAI
Airflow
PySpark

Education

B.E. in Electrical Engineering

M. S. Ramaiah Institute of Technology