I'm Kashif, an AI and backend software engineer building voice-first tools, retrieval applications, computer vision projects, and the APIs behind them.
About
I'm an AI and backend software engineer with hands-on experience building REST APIs, generative AI applications, retrieval systems, voice AI workflows, and computer vision projects. My work spans FastAPI services, Qdrant search, and interfaces that put these systems to use.
I care about practical engineering: clear data flows, verifiable calculations, and tools that solve real problems. From a field-service quoting assistant to sign language recognition, I enjoy taking ideas beyond the prototype.
Experience
FlyRank AI (Remote)
Focus
Turning spoken job descriptions into structured, evidence-backed field-service quotes.
FastAPI services with structured data, persistent workflows, and asynchronous processing.
Qdrant-based information retrieval and computer vision interfaces built for real use.
Projects
Recent work and earlier projects I've built and learned from.
Production Multi-Agent Automation Platform
Building a production-oriented AI agent platform for multi-step business workflows with controlled tools, multi-agent orchestration, and human-in-the-loop approvals. Designing a modular backend for API-driven execution, persistent workflow state, role-based access control, audit logging, and background task processing. Focusing on reliable agent execution, structured outputs, tool permissions, and observable workflows.
Voice-First Field Service Quoting
Built a voice-first quoting system using AssemblyAI Voice Agent and LLM tool calling. FastAPI tools retrieve catalog items, calculate deterministic line-item prices, and save draft quotes in SQLite. A Proof of Hearing workflow links quote evidence to the technician’s original words before sending.
Computer Vision & Deep Learning
Built a DETR and ResNet-50 sign recognition system trained on 15,600 images across 53 classes. Evaluation reached approximately 94.6% mAP@0.50 and 74.1% mAP@0.50:0.95, with PyTorch/OpenCV inference, FastAPI integration, and a React text-to-speech interface.
GenAI Financial Co-pilot
Led a six-member team during a 24-hour AI hackathon to build an LLM-based financial assistant. The earlier prototype explored receipt parsing, financial information retrieval, and SMS alerts.
RAG-Based Clinical Intelligence
Built a Qdrant-backed retrieval application that finds relevant information before generating LLM responses, originally explored for clinical information access.
Data-Driven Environmental Strategy
Explored a GPT-4o Vision workflow for analyzing climate maps and hazard visualizations and generating policy reports.
Stack
The frameworks and ideas I reach for when shipping agentic systems.
AI & Machine Learning
Backend & Web
Data & Databases
Infrastructure & Tools
BS Information Technology, QUEST Nawabshah (2022 to 2026)
Hackathons & AI activities
Certifications & learning
Contact
Hiring, collaborating, or arguing about long-term agent memory, my DMs are open.