Ashank Dsouza
Experienced Engineer
Summary
AI fluency in Claude and Copilot, Founder engineer startup experience, ownership, and ability to move projects from 0 to 1, Enterprise experience, code quality and object-oriented design, Agile methodology experience, Technical mentorship, leadership, and communication skills, Relational database expertise (SQL, PostgreSQL, MySQL), Frontend in React and REST API development experience (NodeJS, Java Spring Boot), Cloud computing and DevOps experience, AI and ML development (RAG, Vector Databases, MCP, ML training, LLM integration)
Education
Bachelor of Science in Computer Science
Master of Science in Software Engineering
Experience
ABC Academy of Music
- Built a low latency frontend using server-side rendering (SSR) of Java Spring Boot REST APIs backed by PostgreSQL.
- Automated a CI/CD pipeline on AWS EC2 to have continuous releases and 99% availability of the site.
- Increased the team’s feature delivery velocity by 75% by introducing the team to an agentic workflow that uses detailed JIRA stories as context for Copilot while leading agile practices.
Tecnotree
- Awarded the “Spotlight Top Performer” cash award for leading high-quality feature delivery in React and Node.js.
- Led development of React UI stepper form with minimal bugs within schedule earning accolades.
PangeaTech
- Founding engineer who built a greenfield ML product from 0 to 1, owning performance, infrastructure, and deployment.
- Distributed logging system: Replaced manual server login and exception trace inspection with a searchable, real-time Kibana dashboard, significantly reducing debugging time for pipeline errors (from 4 hours to real time).
- Scaling data processing: Reduced data processing time by 80% (4 hours to 45 minutes) by redesigning the Python ML backend to distribute workloads across Kubernetes nodes saving customer time.
- Reduced Azure cloud costs 30% with better Azure Virtual Machine utilization.
- Introduced MLflow for ML model version management, logging, testing, and experimentation across 5 algorithms and 100+ experiment runs, enabling reproducible model comparison and faster iteration.
- Used Highcharts and Chart.js in the React frontend app to create 25+ highly customized charts to enhance data visualization for clients powered by data from MongoDB, served by an Express.js API service (Node.js).
PangeaTech
- Engineered a 97% reduction in MySQL database refresh duration (from 8 hours to 15 minutes) by converting synchronous Python to asynchronous Python, saving significant client time.
- Data collection using async API calls throttling the rate to comply with API rate limits. Controlled data fetching and flushing to prevent OOM exception; batching up write requests using SQL bulk insert.
- Maintained MySQL servers at 99% availability and debugged and maintained SQL queries and stored procedures.
- Reduced deployment time by 85% (from 27 to 4 minutes) and deployment failures by containerizing the application with Docker and automating the pipeline via GitHub Actions saving dev time.
Tenerity (a J.P. Morgan and Chase company)
- Reduced network requests by 35% by migrating a microservice from REST to GraphQL, eliminating over-fetching.
- Increased click-through rate by 15% by switching to a new cross-sell itineraries modal developed with Angular.
- Increased test coverage to 90% and ensured compliance with accessibility standards (WCAG) and internationalization.
Projects
Simulation-based system design learning platform (Leetcode for system design)
- Built a simulation engine with TypeScript using prompt-driven development with Claude Code and Copilot.
- Built a custom renderer engine in React for animation using custom React Hooks, Context API, SVG and CSS.
- Managed infrastructure and deployments on Azure VM, including CI/CD pipelines, secure hosting (TLS), SEO for visibility, and site reliability.
Model Context Protocol server to help AI agents better tailor travel recommendations
- Provides JSON-RPC based MCP endpoints to help AI agents provide better travel recommendations across offers.
- Typescript written REST API service which serves tailored recommendations considering location, loyalty points, pagination, client offerings, previous bookings and previous offer clicks.
- Fault-tolerant service with fallbacks on service non-availability, rate limiting to handle excess requests and which points Agentic client to alternate resources on failure.
RAG and LLM based music discovery system
- Built a GenAI-powered music recommender using PostgreSQL + pgvector for embedding-based semantic retrieval, with LLM-driven song summarization and recommendations.
Game written in Python with custom graphics and audio effects
- Received the “Best Paper Award”, a university level award for my paper “Indian Sub-Continent Risk Game”.
Technical Skills
Skills: Full-stack Technologies: Node.js, Express.js, TypeScript, Python, Java (Spring), React, Redux, JavaScript, CSS, HTML, Database Technologies: MySQL, PostgreSQL, MongoDB, SQL, Elasticsearch, DevOps, Infra & Cloud: Docker, GitHub Actions, Git, Linux, Azure (VM, Service Bus, Blob Storage), AWS (EC2)
Awards & Achievements
- Best Paper Award — University
- Spotlight Top Performer — Tecnotree (04/2023)