Prateek Z

Machine Learning Engineer

PUMACY • Fulltime

May 2022 - Present • 4 yrs 3 mos

* Own the ML stack for Plexalytics, an automotive intelligence platform tracking recalls, complaints, technical bulletins, and investigations globally, used by OEM and parts-manufacturer clients to surface safety and quality issues early. * Built and deployed NLP models for summarisation, classification, entity extraction, pattern recognition, and anomaly detection, removing ~40% of irrelevant documents and combining classical ML with LLMs via OpenAI and Anthropic APIs. * Designed and shipped a production RAG service using FastAPI, Qdrant, vLLM, TEI, and Docker. LLM-based intent routing across 5 modes reduced typical analyst research from hours to minutes. * Rebuilt retrieval as a hybrid pipeline combining dense search with deterministic metadata and entity filters, resolving systemic accuracy issues and raising evaluation-set precision to ~90%. * Instrumented ML services with structured logging, tracing, latency, error-rate, and cost metrics, reducing regression identification time from days to under an hour. * Built a CI evaluation harness with curated Q&A and 130+ ingestion regression tests, catching model and retrieval regressions before release. * Optimised inference through batching, caching, and asynchronous jobs, roughly halving median response time under concurrent load while reducing request cost. * Designed graceful degradation, failure isolation, health checks, and versioned rollbacks to prevent model or index failures from becoming outages. * Automated ingestion and synchronisation with scheduled orchestration, validation, and idempotent reruns, removing ~10 hours of manual data operations weekly. * Designed ETL connectors and managed Elasticsearch/SQL infrastructure across client environments, including index design, query optimisation, and capacity planning. * Built React and Power BI dashboards, mentored junior engineers, and worked directly with OEM clients across EU research projects.

Data Engineer

Mobile Premier League • Fulltime

Nov 2021 - Mar 2022 • 4 mos

* Architected and maintained scalable, distributed data pipelines processing event streams from 70M+ users on AWS, powering real-time analytics and reliable data availability across the platform. * Built and optimised high-volume data processing workflows using Databricks and PySpark, tuning partitioning, job configuration, and resource utilisation to improve throughput and reduce processing and storage costs by ~25%. * Owned the analytical data layer, developing complex SQL and Hive workflows that powered daily reporting, analytics, and downstream data products. * Introduced Data Vault architecture to improve data governance, consistency, lineage, and long-term scalability across evolving data sources. * Developed ML-driven workflow automation to reduce manual data operations, improve reliability, and streamline end-to-end data handling. * Designed and integrated data systems across multiple platforms, ensuring reliable end-to-end data flow with minimal downtime and consistent pipeline output. * Established pipeline monitoring for health, data freshness, and operational reliability, working closely with product, analytics, and engineering teams to translate business requirements into scalable data services.

Software Engineer

Dassault Systèmes • Fulltime

Sep 2018 - Nov 2021 • 3 yrs 2 mos

* Delivered end-to-end data analytics and engineering for global enterprise clients across EPC, oil & gas, automotive, and manufacturing, transforming SQL, NoSQL, and flat-file data into decision-ready insights. * Built scalable big data pipelines in Python and Java, integrating and processing data from 25+ disparate source systems into a unified reporting and analytics layer. * Applied machine learning techniques including classification, clustering, pattern analysis, and data modelling to uncover insights and automate analytics workflows across complex enterprise datasets. * Designed executive-facing Tableau and Exalead dashboards, consolidating cross-business KPIs into actionable reporting used by stakeholders across multiple business units. * Improved data quality and reliability within Enovia/PLM ecosystems, translating complex business requirements into technical specifications and production analytics solutions. * Led end-to-end analytics delivery in Agile/SCRUM environments, coordinating cross-functional teams and managing development workflows using JIRA, Git, and Redmine.