Architected a stateful multi-agent workflow using LangGraph, orchestrating Geocoding, Retrieval, and Scoring to deliver structured, reasoning-backed recommendations.

Engineered a two-tier caching strategy (File-system + PostgreSQL) with configurable TTL policies, significantly reducing external API latency and redundant costs.

Built a resilient data pipeline integrating Google Places & Amadeus APIs with automatic multilingual translation and fallback strategies for high-fidelity retrieval.

Architected an Agentic RAG system for Java-to-Python migration, reducing manual refactoring time by ~40% via structured plans and fully typed artifacts.

Engineered multi-turn orchestration with tool-based retrieval and self-correction loops to validate syntax and enforce domain-specific patterns.

Implemented semantic search (ChromaDB) to ground context and a Mem0 persistent memory module to enforce user-specific coding standards (e.g., "prefer Pydantic") across repositories.

Engineered end-to-end pipelines using fastai and PyTorch, deploying production-ready Vision (ResNet) and NLP (Transformer) models to Hugging Face Spaces.

Optimized model convergence by manually implementing core algorithms (SGD, Backpropagation) to diagnose training issues and tune hyperparameters.

Product Innovation: Built the first-ever points-based rewards system for money transfers, collaborating cross-functionally with Design and Product to move from MVP to a high-growth engine (+103% sign-ups).

Infrastructure & Scaling: Migrated core notification services to Java/GQL to streamline our stack and remove architectural bottlenecks (Rust middle-layer removal).

Global Expansion: Engineered the rails for Philippines and Mexico corridors, managing vendor integrations and delegating tasks across the backend team to ensure 100% market coverage.

Engineering Culture: Improved developer experience by introducing automated linting and better error-surfacing tools, while scaling the team through rigorous technical interviewing for backend and infra roles.

Fully owned development on Charger Network Infrastructure Protocols (OCPI), forming and leading a team of 4 engineers.

Designed the data model and significantly contributed to the UI of a Salesforce Contact App. Application launched as part of a new Salesforce Edition aimed at small and medium businesses (SMB).

Collaborated with cross functional teams, architects, PMs and UX Designers across multiple projects.

Mentored and grew junior engineers

Led the migration of internal microservices to GCP by dockerizing and setting up Helm Charts for orchestration and scaling via Kubernetes. Collaborated with the Internal Security Team while minimizing address security concerns.

Improved the code coverage pipeline and dataflows for a distributed Java ETL pipeline that retrieves coverage data from various internal sources and uploads it to GCP BigTable. Added coverage metadata variety and corresponding APIs for retrieval in GCP BigTable. Built using Spring Boot, Apache Camel, and Apache Beam.

Drove releasability initiatives of the team’s microservices. Migrated microservice code from monolithic codebase into Git. Dockerized and set up Terraform scripts to automate the host server creation and application deployment. Successfully reduced the team's release cycle from 3 weeks to on-demand release cadence.