Hi, I'm Aditya
Building full-stack software applications & intelligent AI systems.
Currently interning at Eli Lilly and Company, shipping enterprise GenAI platforms used daily by 200+ employees. Former medical AI researcher (EfficientNet & 3D U-Net). Passionate about distributed systems, vector search, and high-performance engineering.
Technical Skills
Core languages, frameworks, and tools I use to build production systems.
Work Experience
Software Engineering Intern
Eli Lilly and Company
Built and shipped full-stack features (Next.js, React, TypeScript, PostgreSQL, Redis) across an enterprise assistant platform serving 200 internal employees to create and share domain-specific AI assistants. Integrated Veeva Vault via Model Context Protocol (MCP) with SSE streaming, and built admin governance tooling for model whitelisting and per-assistant access control. Resolved the majority of critical/high frontend security vulnerabilities through dependency auditing and npm override strategies.
Research Intern
International Centre for Applied Sciences (ICAS)
Built a hierarchical multi-head EfficientNet-B0 classification pipeline on DERM12345 (12,345 dermoscopic images, 40 subclasses), achieving 98.32% binary accuracy and 0.994 AUC — establishing early benchmarks on a dataset published in Scientific Data (Nature, 2024). Conducted a 6-model systematic study of loss functions (Dice, Tversky, Focal Tversky) on 3D U-Net for BraTS 2020 brain tumor segmentation, achieving Dice scores of 0.77 (Tumor Core) and 0.75 (Enhancing Tumor).
Software Engineer Intern
Recursive Zero
Migrated multiple client websites from vanilla HTML/JS to Astro.js, cutting JavaScript bundle size by ~50% and improving Core Web Vitals and SEO performance across all projects.
Projects
Production systems, vector engines, distributed brokers, and AI pipelines built from scratch.
Mini Vector DB
GitHubvector db from scratch. 99.5% recall.
HNSW vector database engine built from scratch — 99.5% recall@10 on 10K vectors, matching hnswlib and FAISS accuracy.
Mini Kafka
GitHubkafka from scratch. zero deps.
Distributed message broker in Go replicating Kafka's core — append-only logs, partitioned topics, consumer groups, zero dependencies.
Self-Healing RAG
GitHuba RAG that knows when it's wrong
RAG with self-validation and adaptive retry — LLM-as-Judge validates answers, auto-heals with 5 strategies when confidence is low.
RAG Systems
GitHubproduction RAG. the real deal.
Full-stack RAG with hybrid search, recency-aware ranking, SSE streaming, conversation memory, and RAGAS evaluation metrics.
RunCoach AI
GitHubAI yells at me to run faster
Intelligent running app with GPS tracking, real-time AI voice coaching (3 personalities), route mapping, and performance analytics.
Document Intelligence
GitHubthe first RAG build
RAG pipeline over unstructured documents with semantic search and real-time QA.