Built the full-stack MVP at Filmic Technologies: Next.js + Nest.js with Stripe payments, PostgreSQL replication, WebSockets and a CI/CD pipeline stood up from zero. Led two interns and kept it up at 99.99%.
Full-stack media marketplace MVP. Owned backend, led two interns, shipped a production-ready Stripe gateway with Postgres replication, WebSockets and webhook handling in three months.
Shipped a RAG chatbot MVP in six weeks that enabled the founders to apply to accelerator programs with a live demo. Cut response time from 9s to 3s via token-level streaming and reduced hallucination by 40% with source-grounded prompting.
Engineered a hybrid vision-transformer document parser that beat the previous paid solution by 5% (mAP 0.82 @ IoU [0.50:0.95]) and processed 70k+ complex PDFs for downstream RAG pipelines.
Contributed a core pre-processing pipeline to AlibabaResearch's VGT library so the community can run the models on unseen machine-readable PDFs.
Integrated 20+ custom image-processing filters into a Django backend and refactored synchronous pipelines into async microservices via Celery for 1.5x throughput.
Built a hybrid mentor recommendation system serving >250k active users and dropped API latency from 1500ms to under 150ms by redesigning compound indexes improving the response times of all downstream microservices.
Developed an image and video processing pipeline to automate clean-up of user-uploaded images and improve the overall UI of the mobile-app. Reduced 80% manual QA effort.