DEEPAK S MARATHE

Led a 40-member polymath team of STEM experts advancing AGI research systems in collaboration with Google DeepMind.

Directed multidisciplinary AI research pipelines, translating theoretical AI into scalable intelligence systems.

Built and deployed a No-Code, RAG-based AI analytics platform in 175 national languages, enabling non-technical users to run self-service analytics, reducing dependency on data teams by 60% and accelerating decision-making.

Delivered four enterprise pilots (Dawa.ai, Baway, ZeroCostWater, others), processing 100M+ rows, driving 30-40% faster decision cycles, and generating actionable insights across multiple domains.

Architected a secure, scalable full-stack system with AWS (Lambda, EC2, S3), Stripe, Firebase, and GWT-based APIs, cutting infrastructure costs by 25% while maintaining enterprise-grade compliance.

Defined product vision and go-to-market strategy, onboarding the first three enterprise clients and generating $XXK ARR in the first year.

Directed engineering, infrastructure, product, and business development, scaling a 5-member distributed team and achieving MVP launch in under 6 months.

Developed and deployed reporting and analytics APIs (Python, Django, Pandas, MongoDB), reducing query latency by ~40% and improving dashboard responsiveness for internal analytics and client reporting.

Increased unit test coverage to 95% by refactoring legacy code into modular, testable components, reducing production bugs by 25% and accelerating new developer onboarding.

Provided 24/7 production support for critical workflows, maintaining 99.9% uptime for reporting services and ensuring business continuity.

Conducted code reviews and collaborated with data engineering, product, and business teams to deliver high-quality, maintainable solutions aligned with business needs.

Improved analytics platform performance and reliability, enabling faster, data-driven decisions and enhancing stakeholder trust in reporting outputs.

Migrated 50+ on-premise ETL pipelines to AWS Cloud, improving scalability and reducing batch runtime by ~30%, enabling faster analytics and reporting.

Developed Node.js utilities to ensure migration accuracy, zero discrepancies, and boosted stakeholder confidence.

Modernized ETL workflows using Java, Pig, Spark, Hive, and backend validation APIs (Spring), enhancing data integrity and operational reliability.

Implemented deployment automation using Docker, Kubernetes, and Ansible, improving CI/CD efficiency and reducing manual deployment errors by 40%.

Participated in project and capacity planning, Agile ceremonies, and coordinated across engineering, product, and business teams to deliver timely, accurate migrations.

Developed and deployed Doctor Referral Service (Java, Dropwizard, AWS Lambda), enabling seamless specialist referrals and reducing manual referral handling by 40%.

Built Recent Doctors Service (Golang + AWS Lambda) to surface previously consulted doctors, increasing user engagement by 25% and improving repeat teleconsultations.

Created Doctor Categories microservice with MySQL domain modeling, improving search accuracy and reducing average user search time by 30%, enhancing patient experience.

Delivered scalable RESTful APIs on serverless AWS Lambda architecture, ensuring high availability, auto-scaling, and reducing infrastructure costs by 20%.

Led backend design and implementation, collaborating with product, frontend, and operations teams to deliver robust, user-facing features on schedule and aligned with business goals.

Built a self-serve data platform processing ~10M transactions/day, enabling real-time insights and reducing reporting delays by 50% across business units.

Designed and developed Dagger, an open-source Kafka → pluggable sink pipeline, improving ingestion reliability and data flexibility for 15+ downstream services.

Implemented fault tolerance, retries with exponential back off, and monitoring using Datadog, Statsd, and JMX, reducing pipeline failures by 35% and ensuring high availability of streaming analytics.

Managed streaming aggregation pipelines (Kafka, Flink/Yarn, InfluxDB, Grafana) and high-availability Flink clusters on GCP DataProc, increasing real-time analytics throughput by 40%.

Built fraud prevention microservices (wallet suspension, rate limiting) in Java/Spark using Hystrix and Guava, reducing fraudulent transactions by 3% and improving system reliability.

Integrated Voice Capability into QuickBooks Online using Amazon Alexa API within 7 days for hands-free workflow of SMB users, resulting in 5% increase in adoption.

Designed, developed, and deployed a RESTful API for distributed config management using Java, AWS, improving configuration scalability/reliability across 2 products with 15 services.

Increased Dragline's code coverage from 20% to 80% using Sonar, JaCoCo, applying OOP, SOLID principles, and writing tests, resulting in 50% fewer data quality issues and faster ETL migration.

Developed a pluggable mock-data testing framework (Local FS, HDFS, S3, Vertica, Hive), reducing QA cycles by 25%.

Implemented CI/CD pipelines to reduce manual builds and automate deployment using Jenkins, Sonar, improving release efficiency and reliability by 40%.

Built and scaled ETL pipelines from a few GB to ~100 TB volume using HBase, HDFS, Thrift-LZO, MapReduce, and Pig, supporting 800 million active users and enabling data-driven ad targeting.

Developed user profiling and segmentation pipelines for interest-based, retargeting, and persona-targeting campaigns, increasing targeting accuracy by 18% and CTR by 10% across over 100 campaigns.

Built visualization and reporting dashboards with Grafana/Nagios to monitor pipeline health, reducing infrastructure costs by 35%.

Provided 24/7 on-call support with SLAs from 1 to 24 hours for hourly, daily, and monthly data pipelines.

Implemented user data ingestion, curation, enrichment, and caching pipelines, coordinating with Product, BI, Ad-Serving, and User teams to ensure reliable delivery to data centers and online caches.