AI developer building LLM applications in Python: RAG chatbots, Google ADK multi-agent apps, and guardrails that check answers against source data. I write tested, Dockerized FastAPI services with CI. Final-year BS Data Science student at Ziauddin University, Karachi. Open to on-site roles in Karachi and remote roles.
AI Developer (Remote), TechNexus Virtual University · May – Jun 2026 Built FastAPI chatbot back ends for 3 client projects (healthcare, e-commerce, beauty salon) as part of the AI team: RAG over client knowledge bases, intent detection, usage limits, emergency detection and answer safeguards.
Sub Engineer-II (Technical Support), SUPARCO, Karachi · Jul 2022 – Jul 2025 Kept procurement and inspection records, prepared technical reports for senior management, designed PCBs in Altium and tested items against quality standards.
InsightForge: ask questions about an uploaded CSV/Excel file. A deterministic router runs only the Google ADK agents (Gemini) each question needs; a figure and attribution verifier and layered guardrails check every answer. Built by Ameema Rashid. Originally built as a Google ADK capstone project. 552 tests, 94.9% coverage, fuzz-tested on 17 file shapes.
Hoku Health AI: FastAPI health-chatbot back end. Emergency detection runs before any LLM call; dose and diagnosis guards; doctor suggestions from the database; JWT auth with roles; RAG on PostgreSQL with pgvector. Built during my TechNexus internship, published with permission. 707 passing tests (716 collected), 88.5% coverage, mypy strict, CI on PostgreSQL. Found and fixed a silent pgvector search bug.
RRVDXB AI Chatbot: FastAPI shopping assistant with ChromaDB retrieval, a Groq LLM, JWT auth and a price guard that blocks prices not in the given context or tied to the wrong product. Built during my TechNexus internship, published with permission. 360 tests, 94.4% coverage, including 195 database tests on PostgreSQL in CI.
Celine RAG Backend: RAG concierge API for a beauty salon: FAISS retrieval, a Groq LLM, English/French detection, discount rules in code, a CHF price guard and streaming replies. Built during my TechNexus internship, published with permission. 225 tests, 96.7% coverage; retrieval hit@3 0.91 on a small author-written question set.
Spotter Routing (live app): full-stack weather-aware truck routing with React, TypeScript, Django REST Framework and MapLibre. Compares 3 routes and scores weather risk at each checkpoint. Built for Spotter AI's Full Stack technical assessment. 554 tests.
Freight Rate ML: freight rate prediction with scikit-learn and time-based validation; Sep-Oct MAE $98.57 vs $140.53 baseline. Built for Spotter AI's ML Engineer technical assessment.
Bootcamp Buddy: Telegram study bot that answers from lecture slides with citations, using ChromaDB and Groq.
Also (not on GitHub): final-year project, an Outcome-Based Education System (team of 3, ongoing: React, FastAPI, MySQL, scikit-learn), and a Power BI restaurant sales dashboard.
- AI / LLM: Google ADK (multi-agent), Gemini, Groq API, LangChain (langchain-core), RAG, FAISS, ChromaDB, pgvector, sentence-transformers, MCP, guardrails, retrieval evaluation
- Backend: FastAPI, Pydantic, SQLAlchemy 2.0, Alembic, PostgreSQL, MySQL, SQLite, JWT auth, rate limiting, Django REST Framework
- Languages & Frontend: Python, SQL, TypeScript, JavaScript, React, Material-UI, Streamlit
- Data: pandas, NumPy, scikit-learn, Matplotlib, Power BI (DAX), Excel
- Tools: Git, GitHub Actions, Docker, Docker Compose, pytest, ruff, mypy
- AI Engineering Bootcamp, CodeGirls × ConsulNet (Oct – Dec 2026, in progress)
- 5-Day AI Agents Intensive, Google & Kaggle (Jul 2026)