Software engineer with a full-stack, mobile and infrastructure background, plus experience designing and shipping production AI/ML and agentic retrieval systems. Most recently a thesis research engineer at Apple (via AFRY), building agentic retrieval systems that ground an LLM's answers in large, fragmented, structured geospatial datasets. Comfortable across the whole stack: frontend and mobile, backend systems, model architecture, CI/CD and deployment.
Experience
AI/ML Thesis Research Engineer
- Designed and built an LLM-grounded retrieval system for natural-language querying over large, fragmented structured datasets, architecting the LLM as a query planner that routes to the right data partition and pushes predicates down to deterministic execution engines (PostGIS, Neo4j) so unnecessary scans and filters never run.
- Built and benchmarked four retrieval architectures end to end (vector RAG, schema-routed, graph-indexed, and text-to-SQL), cutting inference token cost by 3x across paradigms and identifying routing-index quality as the primary driver of answer accuracy.
- Engineered Neo4j / Cypher knowledge graphs for data lineage and provenance, letting every generated answer trace back to its source records for auditability in enterprise settings.
- Built multi-step agent workflows with LangGraph and an automated evaluation pipeline (deterministic-ground-truth benchmark, LLM-as-judge scoring), using MLflow for experiment tracking and Docker for reproducible runs.
Software Engineer, GenAI (Consultant)
- Built and shipped full-stack GenAI features across web and mobile, turning experimental RAG prototypes into production, user-facing tools.
- Designed retrieval-augmented features end to end, from embedding pipelines and vector search through API integration, wiring LLM backends into responsive React and Next.js frontends.
- Automated data-processing and experimentation workflows in Databricks, shortening model-iteration cycles by roughly 30% so the team could test more approaches per sprint.
- Built Docker-based CI/CD pipelines that automated build, test, and deployment across multiple parallel client engagements.
Software Engineer
- Engineered and shipped 10+ cross-platform mobile apps (native Kotlin / Java and React Native) from concept through App Store and Play Store release across a range of clients.
- Owned the full development cycle on each, from architecture and implementation through testing and release, delivering production apps that held up under real user traffic.
- Built CI/CD pipelines and automated unit / integration testing, cutting release-cycle time by roughly 30% and reducing production defects.
- Integrated third-party REST APIs and worked with cross-functional teams to deliver end-to-end features on schedule.
Education
Master of Science in Data Science · Aug 2024 - June 2026
Bachelor of Science in Computer Science · 2021
Skills
Agentic LLM workflows, RAG, hybrid retrieval and systematic evaluation. Covers the full ML lifecycle, from data preparation to deployment.
Building responsive, production web applications end to end.
Native and cross-platform apps shipped to the App Store and Play Store.
Containerised deployment and CI/CD across client engagements.
Relational, spatial, graph and vector stores.
Data pipelines and experimentation at scale.
Projects
Built a fully client-side semantic search engine over 1M+ arXiv papers that runs entirely in the browser with no inference backend, using Transformers.js and WebGPU with automatic WASM fallback. Engineered binary-quantized embeddings with Hamming-distance retrieval and cross-encoder reranking for fast, accurate search on consumer hardware, running the offline embedding pipeline on a 4-GPU cluster. Served a compact binary index from Cloudflare R2 with IndexedDB caching and offline support via a service worker.