Yo, I'm
Ahnaf.
Come look around, I promise it's not all tech stuff.
About Me
hi, i'm ahnaf. i'm 23. i love building stuff. sometimes it works, sometimes it doesn't.
i grew up in Dubai and moved to Canada when i was 19. studying at the UofM. i never did the co-op program or even worked an internship.
i don't really do things one at a time. usually i have one project with high potential and another one i just personally enjoy working on. right now it's something to do with energy + a really overpowered personal assistant.
if i have an idea i want to build it now. not later, not tomorrow. motivation has a half-life and i don't like wasting it.
The Other Stuff
what i do when i'm not coding.
"the money pit"
416hp of bad decisions ✓
2008 IS-F
5.0L V8 • 416hp
8-speed paddle shift
tap for the mod list →
Did you know... →
My favorite car was a yellow Volkswagen Beetle.
I don't know what
happened either.
🎧 what i'm listening through
Energy Pro 22s. PSB P5s. Hifiman XS.
i buy old speakers and explain why they're better.
"Walk in with confidence"
📌 say something
drop a message, a joke, a hot take.
or just lurk. i respect the lurk. →
"i forgor
TODO: stop researching,
just buy it
(never happens)
I love sunsets.
📍 Dubai, United Arab Emirates
Where I've Worked
Growth / Lead of Product @ EnergyIQ
July 2025 — Present| Remote
- Architect and maintain a multi-model AI system combining structured extraction, content generation, and custom RAG for automated German energy market analysis.
- Manage a distributed engineering team across multiple time zones while owning technical architecture, product direction, and cross-functional communication between business and engineering.
- Design and maintain a deep research engine synthesizing web data and regulatory documents into automated reports, reducing consultant workload by 2-4 hours per customer.
- Developed a meeting intelligence pipeline (extract, route, cluster, aggregate) with token-budget splitting and bin-packing for cross-domain business profiling, processing 534 meetings with a 98% first-run success rate — structured for use in due diligence and succession planning.
- Built an AI sales training system with multiple interaction modes and a programmatic validation harness for response quality.
- Built an invoice extraction pipeline for complex multi-page German energy invoices using a hybrid OCR + LLM approach with BAML for structured output validation, currently at 94% field-level accuracy.
- Implemented dependency parallelization reducing pipeline execution time by 75%.
- Reduced LLM processing costs by 8x through model routing based on task complexity..
Some Things I've Built
Featured Project
Chatbox — OpenAI Feature Integration
Contributed to a medium impact open-source project (26.8k stars, 2.6k forks) by implementing the OpenAI o3 function support into the Chatbox application and interface, enhancing AI interaction capabilities. Designed and developed UI components for adding and removing functions, improving user experience and extending functionality for AI enthusiasts and developers. Integrated new features with existing API calls, ensuring backward compatibility while expanding project capabilities.
- React
- JavaScript
- OpenAI API
Featured Project
CData
Streamlined unstructured data organization by implementing BERT embeddings and cosine similarity analysis, optimized through multi-threaded execution on ARM64 architecture, achieving 35% reduction in processing time. Integrated interactive Plotly dashboards to visualize semantic relationships and embedding clusters, enabling efficient pattern recognition and data-driven decision making. Enhanced data classification workflow by engineering an embedding-based comparison system, potentially reducing analysis times by up to 99% in specific use cases while maintaining data integrity.
- Python
- BERT
- Plotly
Featured Project
Athena Calendar Assistant
Developed Athena, an AI application for conversational Google Calendar management. Internal benchmarking and performance analysis position its capabilities among top AI scheduling tools, particularly in reliably executing complex, multi-step commands derived from natural language (92% interaction efficiency). Built the backend infrastructure (Node.js) utilizing PostgreSQL for persistent storage of user profiles, preferences, and conversation history, ensuring accurate context for the AI. Implemented a multi-layer caching strategy (Redis L2, Node-cache L1) to optimize performance by reducing Google Calendar API latency and ensuring rapid data retrieval. Engineered the natural language processing pipeline using a Large Language Model (LLM) to accurately and safely interpret diverse and context-heavy user commands for calendar actions with 98% accuracy. Architected the AI agent pipeline to perform autonomous planning and sequential invocation of multiple tools (function calling) based on conversational context to fulfill complex, multi-step user requests without any extra input from the user.
- Node.js
- React
- PostgreSQL
- Redis
- Google Calendar API
- LLMs
Other Noteworthy Projects
What's Next?
Get In Touch
My inbox is always open. Whether you have a question or just want to say hi, I'll get back to you for sure!