01
END-TO-END ML PIPELINES
Data preprocessing, feature engineering, Scikit-learn & PyTorch model training, automated validation, and production model serving.
PYTORCH · SCIKIT-LEARN · FASTAPI
KHALID KHAN · AI SOFTWARE ENGINEER | ML ENGINEER | FULL STACK DEVELOPER
PAKISTAN, ISLAMABAD (REMOTE)I build AI systems and production software.
I'm Khalid Khan, an AI Software Engineer and Full Stack Developer focused on building intelligent, production-ready software. I combine Python, PyTorch, Scikit-learn and FastAPI with React, Next.js, TypeScript and modern backend technologies to turn ideas into scalable products.
ML·Python·FastAPI·Next.js·Production
BASHvisitor@khalidkhan.pro — interactive shell80×24
visitor@khalidkhan:~$
try:
TRAINTRAINING — RUN #4721LIVE
EPOCH01/40
LOSS—
VAL LOSS—
LEARNING RATE3.0e-4
GPU MEM87.4%
ETA01:54
▸initializing run #4721 — allocating GPUs…
STAT/010
PRODUCTION ML MODELSSVM 90% · XGBOOST 85% · FASTAPI
STAT/020
AGENTIC & REALTIME APPSSHIFTTAB CLI + VISIONBOARD
STAT/030
FULL-STACK TYPE SAFETYTYPESCRIPT · DRIZZLE · FASTAPI
STAT/040
AVERAGE INFERENCE SLAFASTAPI + ONNX RUNTIME
01 /
Capabilities
systems, models, and product work
02
AGENTIC & LLM SYSTEMS
Autonomous coding agents, function calling, tool execution, MCP integration, and multi-model streaming architectures with Qwen, Gemini, and GPT OSS.
AI SDK · OPENTUI · BUN · HONO
03
MODERN FULL STACK APPS
Next.js 15+ App Router, React 19, TypeScript, Tailwind CSS, Shadcn/UI, and high-performance server actions with zero hydration lag.
NEXT.JS · REACT · TYPESCRIPT · TAILWIND
04
REAL-TIME COLLABORATIVE CANVASES
Multi-user live synchronization, canvas drawing engines, organization spaces, and low-latency state replication with Liveblocks and WebSockets.
LIVEBLOCKS · DRIZZLE · NEON DB
05
DATABASE & DATA MODELING
Relational & vector databases, type-safe migrations with Drizzle ORM, Neon PostgreSQL, MongoDB Mongoose, and Redis caching layers.
POSTGRESQL · NEON · MONGODB · DRIZZLE
06
PRODUCTION-READY APIS
High-throughput REST APIs with FastAPI and Node.js/Hono, automated validation schemas, Better Auth / Clerk auth, and containerized Docker setups.
FASTAPI · HONO · DOCKER · BETTER AUTH
02 /
Selected Work
systems · ml · product · tooling
An AI-powered healthcare platform combining trained Scikit-learn ML models with FastAPI inference services and LLM-assisted medical guidance. Delivers high-accuracy predictions across Heart Disease (SVM ~90%), Diabetes (XGBoost ~85%), Stroke (Logistic Regression ~82%), and Liver Disease (Logistic Regression ~80%), augmented with reasoning by Alibaba Qwen, Gemini, and GPT OSS.
SYSTEM:User → Next.js Dashboard → FastAPI → ML Model → Prediction | User → AI Assistant → LLM Provider → Health Guidance
~90%HEART SVM ACCURACY
~85%DIABETES XGBOOST
4ML MODELS IN PROD
<45msFASTAPI INFERENCE SLA
ROLE — ML MODEL TRAINING · FASTAPI INFERENCE APIS · NEXT.JS & DRIZZLE FULL STACK · LLM ASSISTANT
A terminal-native AI coding agent inspired by Claude Code and Codex designed to bring agentic software development directly into the command line. Features streaming AI chat, model selection, slash commands (/model, /new, /delete, /copy, /clear), OpenTUI terminal interface, and sandboxed file-system tools (reading, writing, searching, creating).
OSSOPEN SOURCE AGENT
CLITERMINAL UI (OPENTUI)
MultiLLM PROVIDER SUPPORT
ROLE — CREATOR · AGENT LOOP ARCHITECTURE · OPENTUI INTERFACE · FILE-SYSTEM TOOLS · HONO & AI SDK
A high-performance real-time collaborative whiteboard allowing multiple users to draw, brainstorm, and work simultaneously with low-latency synchronization. Features interactive drawing tools (freehand, rectangle, circle, eraser, undo/redo), organization workspaces, invitation-based collaboration, Clerk auth, and PostgreSQL on Neon DB with Drizzle ORM.
<25msSYNC LATENCY
MultiSIMULTANEOUS USERS
100%TYPE-SAFE DB (DRIZZLE)
03 /
Stack
tools i use every week
MACHINE LEARNING & AI[08]
PythonPrimary · Daily
PyTorchDeep Learning
Scikit-learnML Models · 90% Acc
XGBoostEnsemble Learning
TransformersHugging Face · NLP
CNNs & RNNsDeep Architectures
Pandas & NumPyData & Features
ONNXModel Deployment
LLMS & AGENTS[06]
AI AgentsShiftTab · Autonomous
AI SDKVercel AI SDK
Tool ExecutionFunction Calling
MCP SupportModel Context Protocol
Multi-ModelQwen · Gemini · GPT
OpenTUITerminal Agent UI
FULL STACK & BACKEND[08]
TypeScriptStrict Typing
FastAPIML REST APIs
Next.jsApp Router · SSR
React.jsReact 19 · Hooks
Node.jsBackend Services
HonoEdge & CLI Backends
Tailwind CSSResponsive UI
REST APIsHigh Throughput
DATABASES & AUTH[07]
PostgreSQLRelational DB
Neon DBServerless Postgres
Drizzle ORMType-Safe Queries
MongoDBMongoose ODM
Better AuthModern Auth
ClerkOrganizations & SSO
JWT & OAuthSession Tokens
04 /
Experience
roles and internships
JUNE 2026 — PRESENT[Remote]
ML Engineering Intern· FlyRank.ai
Working on production-oriented Machine Learning workflows, feature engineering, automated validation pipelines, and developing FastAPI REST APIs for serving models in production. Continuously studying and applying new ML concepts to real-world datasets.
PythonFastAPIMachine LearningFeature EngineeringREST APIs
JUNE 2025 — SEPT 2025[Onsite]
Web Development Intern· Mountain View Tech Park
Built responsive web applications using React.js and TypeScript, developed reusable UI components with Tailwind CSS, integrated REST APIs, and optimized frontend performance across devices.
React.jsTypeScriptTailwind CSSREST APIsPerformance
05 /
Feedback
notes from collaborators
“
Khalid bridges the gap between machine learning and full-stack software engineering seamlessly. He does not just train high-accuracy models—he wraps them in clean FastAPI services and builds responsive, intuitive user interfaces around them.
“
ShiftTab demonstrates a deep understanding of terminal user interfaces, streaming LLMs, and agentic tool execution. Khalid builds developer tools that feel native, responsive, and robust.
“
His execution on PulseAI and VisionBoard shows exceptional full-stack craftsmanship: real-time multi-user synchronization, type-safe database schemas with Drizzle, and clean architecture from day one.
06 /
VIEW ALL ESSAYS (3)Writing & Systems Notes
You will explore in this blog that what is Next.js server actions and how they work with pratical examples
TOPIC — Nextjs · javascript · Reactjs
READ FULL ESSAYLearn more about the typescript such as generics, as const keyword, destructing array and object in typescript and more.
TOPIC — ts · react
READ FULL ESSAYLearn about the TypeScript Utility Types How they modify other types and make life easior for generate the other types.
TOPIC — Typescript
READ FULL ESSAYINTERACTIVE PLAYGROUND
ML Engineering Goodies & Tools
Explore interactive token visualizers, fake psql terminals, visual Git commit trees, and convergence audio synths.
Multi-Model Tokenizer
Visual token parser for GPT-4o, Llama-3, BERT, and raw UTF-8 byte streams.
LAUNCH TOOL
PostgreSQL Terminal ($ psql)
Interactive SQL console to query project tables, schema definitions & metrics.
LAUNCH TOOL
Visual Git Commit Graph
Interactive branch DAG to explore engineering milestones & unified diffs.
LAUNCH TOOL
Loss Audio Synthesizer
Acoustic Web Audio harmonic synthesizer for gradient descent convergence curves.
LAUNCH TOOL