Hi, I'm Aditya
I'm a full stack engineer and aspiring solutions architect who can't leave things alone until I know how they work. I build stuff with equal parts brain and heart, and I think of engineering as architecture, just with fewer hard hats.
To me, building is more than making code work. It's about exploring systems, understanding the problem at hand, and seeing how it shapes the architecture around it. I'm endlessly curious about how people use everyday apps, what makes them stick, and how a codebase evolves as it scales.
Building
Hivemind
A context engine for your agents. One memory for every AI you use.
LLMs are stateless, and your agents have the memory of a goldfish. You explain your stack to Claude Code, then to Cursor, then to ChatGPT, and tomorrow you do it all again. Hivemind is the context engine I'm building so you only say it once: one memory layer, built on SOTA retrieval techniques, sitting under every agent you use and handing each one just the context it needs. Agentic infra, shipped as a SaaS.
For the retrieval nerds: hybrid search, rank fusion and a cross-encoder rerank, then the winners get packed into your token budget like a carry-on. Nothing relevant? It sends nothing. A dashboard shows you what the model actually saw and what got cut, and it plugs in wherever you already work: MCP for 13 coding tools, a browser extension for chat sites, GitHub and Notion sync, and an SDK for your own agents.
- right on LongMemEval's single-session questions (oracle split). Never tuned on it.
- 98.4%
- the best published score on MemoryAgentBench's conflicting-facts, multi-hop split. 10% vs 6%, 100 questions.
- 1.7x
- fewer tokens than dumping the whole chat in. 465 a question vs 18,853 on LoCoMo.
- 40x
- Tell one tool, ask another. 20 for 20 across chat, MCP and CLI.
- 20/20
Writing
All postsOpen source work
All projectsClipper
An AI agent that chops long videos into social-ready clips. Gemini 2.5 Pro does the watching, FFMPEG does the cutting.
Next.js
PostgreSQL
FastAPI
Gemini 2.5 Pro
Tailwind + ShadCN
Gitbuddy
A crew of agents that handles the repo chores nobody wants: docs, Dockerfiles, and commit history.
- Next.js
Supabase
Langchain- Gemini 2.5 Flash + 2.5 Pro
- Tailwind + ShadCN
GraphRAG Chat
A chatbot that scrapes the web, builds a knowledge graph, and answers the hard questions with receipts. Serverless on GCP.
- Next.js
- FastAPI
Spanner Graph DB- Langchain
- Gemini 2.5 Flash
Finchat
Ask it about a stock and it answers in real time, charts included. RAG and GPT-4o under the hood.
Python- Fast API
GPT-4o, Azure OpenAI- LangChain, LangGraph
Pandas
Where I've been
Associate Software Engineer at Enercare
Full stack, all the way down: TypeScript, Nest.js, Azure, React. Mostly I'm dragging legacy systems into a reliable, event-driven future.
Software Development Mentee at Architech
Built AI customer support automation with LangGraph and ReAct, got the LLM giving sharper answers with semantic search (OpenAI embeddings, pgvector, HNSW), and set up CI/CD so our Python and React apps stopped needing a babysitter.
Software Development Intern at J&M Group
Designed a location-aware hiring system with PostGIS, shipped pieces of a Next.js PWA job board, and kept our self-hosted Docker Swarm alive and mostly happy.
MS Information Technology at Humber Polytechnic
BS Computer Applications at GLS University
My toolbox
- Frontend
TypeScript
JavaScript
React, React Native- Next.js
TanStack Query
Redux- Tailwind
Storybook
- Backend
Node
Nest.js
Bun
Deno
Spring Boot- FastAPI
Kafka
WebSockets
Dapr
Prisma
TypeORM
- Data
- PostgreSQL
MySQL
MS SQL
MongoDB
Redis
Firebase- Supabase
Convex
- Cloud
Azure, Azure DevOps
GCP
Cloudflare
Oracle
Vercel
Docker
GitHub Actions
Bash
OpenTelemetry
Datadog
Sentry
- AI
Claude
Codex
OpenCode
Ollama- Vercel AI SDK
- LangChain
- Vertex AI
MCP
TensorFlow
- Testing
Jest
Playwright
