I like turning small annoyances into usable software. These days I build production software at CarbonTeq; outside work, I keep following questions into web products, AI workflows, and small systems projects.
About
A paper, a confusing workflow, or a protocol I have only met through a library is usually enough to send me into the system. I follow the friction until I can make something useful from it.
Now
Lahore
CarbonTeq
Building production software across product interfaces, backend systems, and practical AI workflows.
2025
Islamabad
IVY Interactive Solutions
Worked on automated project planning, real-time analytics for early risk detection, deployments, and technical documentation.
2024
Lahore
CarbonTeq
Built RAG and predictive-analytics work, then automated model delivery through CI/CD.
2022
Lahore
CarbonTeq
Developed authenticated REST APIs with JWT and MongoDB, and built a DNS client to work directly with the protocol.
Now
Lahore
CarbonTeq
Building production software across product interfaces, backend systems, and practical AI workflows.
2025
Islamabad
IVY Interactive Solutions
Worked on automated project planning, real-time analytics for early risk detection, deployments, and technical documentation.
2024
Lahore
CarbonTeq
Built RAG and predictive-analytics work, then automated model delivery through CI/CD.
2022
Lahore
CarbonTeq
Developed authenticated REST APIs with JWT and MongoDB, and built a DNS client to work directly with the protocol.
Capabilities
React, Next.js, TypeScript, responsive UI
Node.js, Express, Flask, REST, GraphQL
AI agents, RAG, LangChain, Gemini, OpenAI APIs
Computer vision, deep learning, multimodal models, PyTorch, TensorFlow
PostgreSQL, MongoDB, Docker, CI/CD, monitoring
React, Next.js, TypeScript, responsive UI
Node.js, Express, Flask, REST, GraphQL
AI agents, RAG, LangChain, Gemini, OpenAI APIs
Computer vision, deep learning, multimodal models, PyTorch, TensorFlow
PostgreSQL, MongoDB, Docker, CI/CD, monitoring
Selected work
Reading papers was starting to feel like a tax on curiosity. Before I could decide whether a paper deserved an hour, I had to decode the abstract, skim the method, and hunt for the actual finding. I built Briefly to make that first pass faster.
University information was scattered across portals, handbooks, faculty pages, and files that students only found after asking the right person. For my final-year project, I built CampusHive: a multi-agent university assistant that routes each question to the part of the system equipped to answer it.
I take photos too, and I wanted a small place for them that did not behave like a social feed. Pixel Arrays is intentionally quiet: the interface steps back and lets the photographs carry the page.
I wanted to understand what changes when blockchain transactions carry executable ML work instead of only values. I built a Go master-peer network where peers run Python models, validate their outputs, compete on proof of work, and broadcast the winning block.
Master
Peer 01
Peer 02
The first valid block propagates. Competing mining work is cancelled.
More projects
I wanted to see what banking would feel like if the interface was a conversation instead of a maze of tables. BTC Conversa explores chat as a way to explain transactions, patterns, and next steps.
PDFs hold useful information, but asking them direct questions is awkward. I built the retrieval path myself so answers could be generated from the document rather than from an unconstrained prompt.
I built this because I did not want DNS to stay a black box. The point was to construct the packets, send them over sockets, and inspect what a resolver is actually doing.
I wanted to understand interpreters by making one small enough to hold in my head. Forth was a good fit because its stack model is simple but still forces careful thinking about parsing and execution.
Contact
I am most useful when a product has a real problem behind it: a workflow that is too slow, information that is hard to reach, or a system that needs to become easier to trust. Send the rough context first.