I build AI-powered applications using Python, APIs, retrieval-augmented generation (RAG), and agentic workflows. My technical focus is making AI applications more grounded, testable, and reliable.
I'm strengthening my ability to take AI applications from prototype to tested, deployable software, with a long-term interest in aerospace, satellite, and autonomous-systems software.
📍 Pakistan · Open to relevant remote and international opportunities.
- 🏆 IBM Granite Generative AI Hackathon — Winner, Team PolyEns.
- 🧩 Harvard CS50x Puzzle Day 2025 — 1st Place, reported score: 9/9.
- 🎓 Stanford Code in Place — Section Leader, 2025 and 2026.
- 🚀 BICON Startup Incubation — Selected for final-year project. Supporting link
- 🔬 AlgoVerse Research Program (EOE.ai × Omdena) — Accepted with a reported 45% merit scholarship; RAG/NLP research.
- 🌍 International hackathons: 17 events and 14 submissions, as reported on my LabLab profile.
My focus is the engineering side of applied AI: turning model capabilities into applications that solve practical problems.
I enjoy working across the application stack, from API development and data retrieval to LLM orchestration, evaluation, and user-facing interfaces.
- Applied AI: Building LLM-powered applications for practical workflows.
- Backend engineering: Developing Python services, REST APIs, and data-backed applications.
- RAG and information retrieval: Connecting generated answers to relevant source material.
- AI evaluation and reliability: Exploring groundedness checks, citation validation, critic/verifier patterns, and uncertainty handling.
- Agentic workflows: Designing multi-step AI workflows with explicit responsibilities.
- Product-oriented engineering: Connecting technical implementation to user needs, observable behavior, and measurable outcomes.
My engineering goal is to strengthen the full lifecycle of AI applications:
Prototype → Test → Deploy → Evaluate → Improve
I'm exploring aerospace, satellite, and autonomous-systems software, particularly engineering workflows where information retrieval, evidence traceability, and reliable AI-assisted analysis may be useful.
This is an area of focused learning and project exploration—not a claim of aerospace industry experience.
These projects demonstrate different aspects of my applied AI work. Each description explains the intended functionality and implementation areas without assuming production readiness.
Exploring how to assess whether generated answers are supported by retrieved evidence.
RAG applications can produce plausible answers that are not adequately supported by their source documents. ComplianceRAG explores a retrieval-and-generation pipeline with an additional groundedness critique step.
What it demonstrates
- Document ingestion and semantic retrieval.
- Retrieval using FAISS and sentence embeddings.
- LLM-based answer generation.
- Sentence-level groundedness critique and citation validation.
- Identification of questionable outputs for further review.
Stack: Python · Gradio · FAISS · sentence-transformers · Hugging Face Inference API · PyPDF
🏆 IBM Granite Generative AI Hackathon Winner — Team PolyEns
A multi-agent approach to evaluating sales leads by considering both opportunity and risk.
The workflow uses agents with different responsibilities to assess a lead, surface potential concerns, and reconcile findings into a combined assessment informed by historical deal context.
What it demonstrates
- Multi-agent orchestration and role separation.
- Structured reasoning across competing perspectives.
- Backend API development.
- An AI-powered workflow connected to a web application.
Stack: Next.js · FastAPI · IBM Watson AI · IBM Granite
An LLM-powered application designed to help candidates analyze job requirements and organize career-preparation tasks.
The application brings related workflows into one interface, including job-description analysis, skill-gap identification, resume assistance, cover-letter drafting, interview preparation, and career-related chat.
What it demonstrates
- LLM integration in a user-facing application.
- Multi-step workflows for career-related tasks.
- Structured outputs and application state management.
- Rapid development of an interactive AI application.
Stack: Python · Streamlit · Groq API · LLaMA 3.1 · Hugging Face Spaces
More work: SafeLite
A modular AI system exploring the separation of planning, safety reasoning, execution, simulation, and evaluation.
Areas of interest include explicit safety checks, component-level testing, controlled simulation, and investigating system behavior under different conditions.
Stack: Python · FastAPI · LangChain · Custom evaluation harness
| Area | Technologies and concepts |
|---|---|
| Languages | Python, SQL, TypeScript |
| Backend engineering | FastAPI, REST APIs |
| Data and persistence | PostgreSQL, SQLAlchemy, Pandas |
| LLM applications | Groq, LLaMA, IBM Granite, IBM Watson AI, Hugging Face Inference API, OpenAI APIs |
| RAG and retrieval | FAISS, sentence-transformers, embeddings, LangChain, PyPDF |
| Agentic workflows | Multi-agent workflows, critic/verifier patterns, CrewAI |
| Frontend and prototyping | Next.js, Streamlit, Gradio |
| Deployment and demos | Hugging Face Spaces, Streamlit Cloud |
This list reflects technologies and concepts I've worked with across projects. It does not imply equal depth or production experience in every item.
Build for a real problem. Start with the problem and intended user, not a framework or model.
Ground answers in evidence. Retrieval quality, source relevance, and context quality matter alongside generation.
Evaluate instead of assuming. A feature existing in code does not establish that it works reliably.
Handle uncertainty explicitly. Questionable outputs should be identified rather than presented as unquestionable facts.
Make systems understandable. Clear architecture, documentation, limitations, and reproducible steps help others assess and improve a project.
Engineer beyond the prototype. Tests, failure handling, security, observability, and deployment matter when applications need to work consistently.
Keep learning through implementation. I value building, testing, and improving working systems over collecting tools for their own sake.
My current priority is to strengthen the engineering practices that help turn AI prototypes into dependable software:
- Automated testing and regression evaluation.
- Backend design, data modeling, and API reliability.
- Deployment, logging, and observability.
- RAG evaluation, groundedness, and failure analysis.
- Understanding aerospace engineering workflows and their information needs.
My long-term goal is to combine applied AI engineering, software reliability, and domain understanding to build useful systems with measurable value.
I'm interested in opportunities involving Applied AI, AI software engineering, LLM applications, Python backend systems, RAG, agentic workflows, and AI evaluation.
I'm especially interested in teams that value sound engineering, practical problem-solving, evidence-based iteration, and learning from real users.
If you're building AI-powered products or working on challenging software engineering problems, I'd be glad to connect.
- Email: faraz.outreach8@gmail.com
- LinkedIn: linkedin.com/in/fm61
- GitHub: github.com/faraz-369
- Hugging Face: huggingface.co/Faraz618
- LabLab: lablab.ai/u/@Faraz_Mubeen
Create with curiosity. Build with responsibility. Contribute with purpose.