# Yaseen Khatib — Senior Full-Stack AI Engineer

> Yaseen Nurmahammad Khatib builds and ships autonomous AI products solo — Agentic RAG
> pipelines, LLM orchestration, and the MERN systems they run on. Five
> production products in twelve months; open to Lead/Senior roles —
> remote, hybrid, or on-site.
> Base: Hyderabad, India (IST). Contact: contact@streamerosai.com.

## Canonical pages

- [About (canonical bio)](https://yaseenkhatib.streamerosai.com/about/)
- [Products — architecture teardowns](https://yaseenkhatib.streamerosai.com/products/)
- [Interview brief for engineering leadership](https://yaseenkhatib.streamerosai.com/interview/)
- [Infrastructure & Resiliency Lab (FinOps simulator, chaos toggle)](https://yaseenkhatib.streamerosai.com/sandbox/)
- [Tooling stack](https://yaseenkhatib.streamerosai.com/uses/)
- [Machine-readable profile (JSON)](https://yaseenkhatib.streamerosai.com/ai-briefing.json)

## Final year projects for Indian students (BCA / MCA / B.Tech)

> Final year projects built end to end for the student — code, database,
> report, diagrams, deck and deployment. The student writes none of it. Any
> idea can be commissioned, not only the listed ones: an approved problem
> statement or the student's own concept, in any stack, quoted
> ₹10,000–₹50,000
> by scope.
> After delivery there are daily live
> Google Meet sessions after 8:00 PM IST where
> the code is explained line by line, ending in a mock viva, so the student can
> defend work they did not write.
> Payment is direct (UPI / bank transfer) in monthly installments — there is no
> checkout, gateway or subscription. Each listing is capped to a small number of
> students per college per year.

- [Catalog — all final year projects](https://yaseenkhatib.streamerosai.com/final-year-projects/)
- [Custom project built to your problem statement](https://yaseenkhatib.streamerosai.com/final-year-projects/custom/)
- [For colleges — workshops and cohort mentoring](https://yaseenkhatib.streamerosai.com/final-year-projects/colleges/)
- [Terms — payment, delivery, cancellation](https://yaseenkhatib.streamerosai.com/final-year-projects/terms/)
- [Free tool: timeline planner — usable weeks before a submission date](https://yaseenkhatib.streamerosai.com/final-year-projects/planner/)
- [Free tool: viva question bank — 83 questions, filterable, with what each one tests](https://yaseenkhatib.streamerosai.com/final-year-projects/question-bank/)
- [Final Year Project Viva Questions (And How To Answer Them)](https://yaseenkhatib.streamerosai.com/final-year-projects/guides/final-year-project-viva-questions/): The questions panels really ask in a final year project viva, grouped by what they are testing, with the reasoning behind a good answer. Written by an engineer who builds these systems for a living.
- [Final Year Project Synopsis Format — Section By Section](https://yaseenkhatib.streamerosai.com/final-year-projects/guides/final-year-project-synopsis-format/): How to write a final year project synopsis that gets approved: what each section needs, the standard structure, the right length, and the mistakes that get a synopsis sent back.
- [Final Year Project Presentation & PPT — Slide By Slide](https://yaseenkhatib.streamerosai.com/final-year-projects/guides/final-year-project-presentation-ppt/): How to build and deliver the final year project PPT for your viva: the slide-by-slide structure, how many slides, what examiners look at, demo tips, and handling questions.
- [How To Deploy Your Final Year Project For Free (Live In The Viva)](https://yaseenkhatib.streamerosai.com/final-year-projects/guides/how-to-deploy-your-final-year-project-free/): Get your final year project live on a real URL for free — frontend on Vercel or Netlify, backend on Render, MongoDB Atlas, secrets done right, and a localhost fallback if the wifi dies.
- [BCA vs MCA vs B.Tech Final Year Projects — What Examiners Expect](https://yaseenkhatib.streamerosai.com/final-year-projects/guides/bca-vs-mca-vs-btech-final-year-projects/): How final year project expectations differ across BCA, MCA and B.Tech, plus B.Sc IT, Diploma and M.Tech — the depth panels want, whether you need a research angle, and how to pick.
- [MCA Final Year Project Topics (2026–27) With Stack And Difficulty](https://yaseenkhatib.streamerosai.com/final-year-projects/guides/mca-final-year-project-topics/): Twenty MCA final year project ideas grouped by domain, each with the stack it needs, how hard it really is, and the viva question it will attract. Written by an engineer who builds these systems professionally.
- [BCA Final Year Project Topics You Can Actually Finish](https://yaseenkhatib.streamerosai.com/final-year-projects/guides/bca-final-year-project-topics/): Sixteen BCA project ideas with the stack each needs and an honest difficulty rating, chosen for projects that can be completed and documented in the time a BCA student actually has.
- [B.Tech CSE Final Year Project Topics That Survive Review Panels](https://yaseenkhatib.streamerosai.com/final-year-projects/guides/btech-cse-final-year-project-topics/): B.Tech final year project ideas chosen for team work and staged reviews — each with the stack, honest difficulty, how to split it across a team, and the question the panel will open with.
- [Final Year Project Report Format — Chapter By Chapter](https://yaseenkhatib.streamerosai.com/final-year-projects/guides/final-year-project-report-format/): The standard structure of a final year project report for BCA, MCA and B.Tech, what belongs in each chapter, the diagrams you need, and the mistakes that cost marks at submission.
- [How To Choose A Final Year Project (Without Regretting It In March)](https://yaseenkhatib.streamerosai.com/final-year-projects/guides/how-to-choose-a-final-year-project/): A practical method for picking a final year project: how to size it against the weeks you actually have, how to avoid duplicating your batch, and the questions to ask your guide before you build anything.
- [E-Commerce Platform — MERN stack with a real cart & checkout](https://yaseenkhatib.streamerosai.com/final-year-projects/ecommerce-mern/): Most e-commerce projects are a product grid with a Buy button that does nothing, and they collapse the moment an examiner asks what happens when two people buy the last item. This one keeps the cart on the server so it survives refresh and login, decrements stock so it can never oversell, and runs a real Razorpay test-mode checkout that produces an order you can open in an admin panel. It behaves like a shop, not a screenshot. Suits BCA, MCA, B.Tech, B.Sc IT. Stack: React, Node.js, Express, MongoDB, Redux Toolkit, JWT, Razorpay (test mode). From ₹5,500.
- [Real-Time Chat App — MERN + Socket.IO with presence](https://yaseenkhatib.streamerosai.com/final-year-projects/realtime-chat-mern/): Most chat projects poll an API on a timer and call it real-time; they lag, they duplicate messages, and they fall apart the moment two people type at once. This one uses Socket.IO for a genuine push connection — one-to-one and group chats, typing indicators, online and last-seen presence, delivery and read receipts, and history that persists and paginates. It survives the question every panel asks: how do you know a message actually arrived, and in the right order? Suits BCA, MCA, B.Tech, B.Sc IT. Stack: React, Node.js, Express, Socket.IO, MongoDB, JWT, Zustand. From ₹5,500.
- [Recommendation System — hybrid collaborative + content filtering](https://yaseenkhatib.streamerosai.com/final-year-projects/recommendation-system-ml/): Most recommender projects hardcode a similarity matrix, demo on the same three users, and freeze the moment an examiner logs in as someone new. This one builds collaborative filtering with matrix factorization, adds a content-based model for cold-start, blends the two, and shows why each item was recommended. It reports precision and recall honestly rather than claiming it is always right, and it answers the question that sinks most of these: what do you recommend to a user with no history? Suits MCA, B.Tech, M.Tech. Stack: Python, pandas, scikit-learn, Surprise, FastAPI, React, PostgreSQL. From ₹6,000.
- [Fake News Detection — an NLP classifier that shows its reasoning](https://yaseenkhatib.streamerosai.com/final-year-projects/fake-news-detection-ml/): Most fake-news projects are a pasted Kaggle notebook that reports 99% accuracy and cannot explain a single prediction. This one builds the NLP pipeline properly — cleaning, tokenizing, TF-IDF, a trained classifier, with an optional transformer upgrade — reports real metrics with a confusion matrix, and surfaces the exact words that pushed a decision. It is built to answer the one question that ends most of these vivas honestly: what does your model get wrong, and why? Suits MCA, B.Tech, M.Tech. Stack: Python, scikit-learn, NLTK, TF-IDF, FastAPI, React. From ₹6,000.
- [Plant Disease Detection — a CNN that works on real leaf photos](https://yaseenkhatib.streamerosai.com/final-year-projects/plant-disease-detection/): Most computer-vision projects demo on an image straight from the training set and fall apart on a photo taken in the exam hall. This one trains a CNN with transfer learning on a leaf-disease dataset, runs on a picture snapped in class, returns the diagnosis with a confidence score and the top three possibilities, and maps that to a plain treatment note. It handles the awkward case honestly — a blurry photo or a leaf it has never seen returns low confidence, not a confident wrong answer. Suits B.Tech, M.Tech, MCA. Stack: Python, TensorFlow / Keras, CNN (transfer learning), FastAPI, React. From ₹6,000.
- [Blood Bank Management System — MERN with donor matching](https://yaseenkhatib.streamerosai.com/final-year-projects/blood-bank-management/): Most blood-bank projects are a table of donor names with no logic behind them, and they stall the moment an examiner asks which group can donate to which. This one manages donors, inventory by blood group, and requests — matching a request to eligible donors using real group compatibility and location, decrementing stock every time a unit is issued, and keeping an audit trail of where each unit went. It is a clean, honest database project that actually enforces its own rules. Suits Diploma, BCA, B.Sc IT. Stack: React, Node.js, Express, MongoDB, JWT. From ₹4,000.
- [AI College Assistant — Agentic RAG over your college documents](https://yaseenkhatib.streamerosai.com/final-year-projects/ai-college-assistant/): Most student chatbots are a wrapper around an LLM and fall apart the moment an examiner asks where the answer came from. This one ingests real college documents — syllabus, exam rules, fee circulars — chunks and embeds them into a vector store, retrieves the relevant passages, and answers only from what it retrieved, with a citation back to the source page. It is the project that survives the hardest viva question: how do you know it is not making it up? Suits MCA, B.Tech, M.Tech. Stack: Python, FastAPI, LangChain, ChromaDB, Sentence Transformers, React, PostgreSQL. From ₹6,000.
- [Hospital Management System (MERN)](https://yaseenkhatib.streamerosai.com/final-year-projects/hospital-management-mern/): The most-attempted project in Indian colleges, built the way it should be. Separate portals for patients, doctors, and admin; JWT auth with role guards on every route; appointment booking that actually detects slot collisions instead of overwriting them; and prescriptions and billing tied to real MongoDB relations. The marks in this project are in the schema design and the auth flow, so that is where the sessions spend their time. Suits BCA, MCA, B.Tech, B.Sc IT. Stack: React, Node.js, Express, MongoDB, Mongoose, JWT, Tailwind CSS. From ₹4,500.
- [Smart Expense Tracker — offline-first mobile app](https://yaseenkhatib.streamerosai.com/final-year-projects/expense-tracker-mobile/): Most mobile projects are a form over an API and die when the campus wifi drops during the demo. This one stores everything in on-device SQLite, works fully offline, and syncs when a connection returns. Expenses are auto-categorised from the merchant text, budgets raise alerts before they are breached, and the charts are computed from real queries rather than hardcoded arrays. It installs on the examiner's phone. Suits BCA, MCA, B.Tech, B.Sc IT. Stack: React Native (Expo), TypeScript, SQLite, Zustand, Victory Native, Node.js, Express. From ₹5,000.
- [AI Resume Screening & Job Match System](https://yaseenkhatib.streamerosai.com/final-year-projects/resume-screening-ml/): Upload a job description and a folder of resumes; the system parses each PDF, extracts skills and experience, scores the match, and ranks the candidates. The part that earns marks is explainability — every score breaks down into the skills matched, the skills missing, and the experience gap, so it is a model you can defend rather than a black box that prints a number. Suits MCA, B.Tech, BCA, M.Tech. Stack: Python, Flask, scikit-learn, spaCy, TF-IDF, PyPDF2, MySQL, Bootstrap. From ₹4,500.
- [Face Recognition Attendance System](https://yaseenkhatib.streamerosai.com/final-year-projects/face-recognition-attendance/): A camera-based attendance system that enrolls students from a few photos, recognises faces in a live feed, and writes attendance to a database with a timestamp. It handles the two things a demo version never does: rejecting an unknown face instead of guessing the closest match, and refusing to mark the same student twice in one session. Examiners always ask how the recognition works — the sessions make sure you can draw it on the board. Suits BCA, B.Tech, MCA, Diploma. Stack: Python, OpenCV, face-recognition (dlib), Flask, SQLite, NumPy. From ₹4,000.

## Founder's Log — vision series

- [How I Went From My First Dev Job to Building AI Systems](https://yaseenkhatib.streamerosai.com/blog/my-journey-web-developer-to-ai-engineer/): Five years, honestly told: an ASP.NET healthcare backend in 2021, full-stack MERN through the ChatGPT shift, nights of embeddings homework, and finally shipping five AI products solo. The journey, without the gloss.
- [I Shipped 5 Products in 12 Months — Solo, Unfunded, and Faster Than Most Teams Ship One](https://yaseenkhatib.streamerosai.com/blog/shipped-5-products-solo-12-months/): Five production systems — a Rust desktop cockpit, a workflow engine, a local-first AI finance agent, and two autonomous pipelines — designed, built, and shipped by one engineer. Here's the operating model.
- [Vision Over Syntax: I Design the Entire Product in My Head Before the First Commit](https://yaseenkhatib.streamerosai.com/blog/vision-over-syntax-architecture-first/): Syntax is solved; vision is not. How architecture-first thinking — data contracts, trust boundaries, and deliberate refusals — is the real engineering skill in the AI era, shown through shipped products.
- [One Architect + Claude + MCP = A Full Engineering Squad: The Operating Model I'd Install at Your Company](https://yaseenkhatib.streamerosai.com/blog/one-architect-claude-mcp-full-squad/): The exact three-layer operating model — architect, Claude Code, MCP-wired systems — that lets one engineer ship like a squad, and the 30-day plan for installing it inside a real company.
- [Your Company Is Overpaying for AI by 80% — I Built a Live Simulator That Proves It](https://yaseenkhatib.streamerosai.com/blog/ai-finops-playbook-stop-burning-money/): Most production AI traffic is simple or repetitive, yet it's billed at frontier-model prices. The cache → flash → frontier cascade that fixes the economics, with a live interactive cost simulator.
- [My Networking Runs Itself: The LinkedIn Pipeline That Grew My Reach While I Slept](https://yaseenkhatib.streamerosai.com/blog/linkedin-pipeline-job-search-runs-itself/): How I engineered consistency: an autonomous GitHub Actions pipeline that turns real work into scheduled LinkedIn posts, tracks state in Git, costs $0 to run — and what the same pattern automates inside a company.
- [The $0 Content Engine: My Blog Writes, Commits, and Deploys Itself Twice a Week](https://yaseenkhatib.streamerosai.com/blog/zero-dollar-content-engine/): A cron-scheduled pipeline drafts articles with Gemini, writes MDX, commits to Git, and redeploys the static site — no CMS, no server, no database, no bill. The architecture of refusing infrastructure.
- [I Built an AI Money Agent That's Structurally Incapable of Touching the Money](https://yaseenkhatib.streamerosai.com/blog/sable-ai-agent-never-touches-money/): Sable is a local-first AI financial agent: all data in on-device SQLite, no cloud backend, and a Review & Confirm boundary where the model proposes but only a human commits. The trust architecture, explained.
- [Everyone Said 'Just Use Electron.' I Wrote streamerOS in Rust — Here's What That Bought](https://yaseenkhatib.streamerosai.com/blog/streameros-rust-over-electron/): streamerOS runs on the most contested consumer PC there is — game, encoder, OBS, chat. Why a Rust + Tauri core, on-device chat ingestion, and a MessagePack IPC bridge were the product decision, not a preference.
- [The 94% Decision: One Architecture Call That Made IntegrateX Feel Instant](https://yaseenkhatib.streamerosai.com/blog/the-94-percent-decision-integratex/): Persisting React Flow's UI objects nearly sank the product. Drawing a serialization boundary between view and domain cut payloads 94% — the reasoning behind the call, and how to audit your own product for it.
- [Hire My Head, Not My Hands: What 5 Solo Products Prove That a Résumé Can't](https://yaseenkhatib.streamerosai.com/blog/hire-my-head-not-my-hands/): AI made hands abundant; heads stayed scarce. The closing argument of the Founder's Log series — the architectural judgment pattern across five shipped products, addressed to the person deciding whether to interview me.

## Recent articles

- [Agentic Workflows vs. Copilots: Autocomplete vs. Autonomy](https://yaseenkhatib.streamerosai.com/blog/agentic-workflows-vs-copilots/): A copilot completes your line; an agent executes your task. Why the difference is categorical — not incremental — and why it separates a 10% speedup from a 10x one.
- [Custom Serialization Adapters: 94% Payload Compression](https://yaseenkhatib.streamerosai.com/blog/custom-serialization-adapters/): Rich UI objects make terrible database records. The Serialization Adapter pattern that separates render model from transport record — and cut IntegrateX payloads by 94%.
- [Vector Embeddings in Production: Custom Search for RAG](https://yaseenkhatib.streamerosai.com/blog/vector-embeddings-in-production/): Demo vector search is one query; production RAG is a pipeline. Chunking strategy, metadata-filtered hybrid search, and index freshness — the architecture behind the Police RAG system.
- [Redis Caching Strategies That Don't Serve Stale Data](https://yaseenkhatib.streamerosai.com/blog/advanced-redis-caching-strategies/): On a busy portal I learned Redis caching strategies live or die on details: one interceptor boundary, cache-aside vs write-through, TTL jitter, stampede guards.
- [WebSocket Telemetry at Scale: Real-Time Streaming Architectures](https://yaseenkhatib.streamerosai.com/blog/websocket-telemetry-at-scale/): A single WebSocket server is a weekend project; streaming to thousands across instances is architecture. Redis pub/sub backplanes, room-scoped subscriptions, and source-side coalescing.
- [MongoDB Aggregation Pipelines: Stage Order Is the Win](https://yaseenkhatib.streamerosai.com/blog/optimizing-mongodb-aggregation/): Profiling a clinical workflow API taught me MongoDB aggregation lives or dies on stage order: $match first on indexes, $lookup treated as the join it is, then read explain.
- [Budgeting the Claude Context Window Before It Truncates You](https://yaseenkhatib.streamerosai.com/blog/mastering-claude-architecture-context-windows-output-limits/): The Claude context window is one token budget shared by prompt, tools, thinking, and reply. I count tokens and guard it in code so a reply never truncates.
- [Role-Based Access Control Patterns for Node.js APIs](https://yaseenkhatib.streamerosai.com/blog/role-based-access-control-patterns/): Authorization bugs don't crash — they let the wrong person succeed. Enforcing RBAC at the middleware layer with JWT claims, fine-grained permissions, and deny-by-default.
- [XML-Tag Structural Prompting: Claude's Deterministic Shell](https://yaseenkhatib.streamerosai.com/blog/xml-tag-structural-prompting-deterministic-shell/): Flat prompts make Claude guess which text is a command, context, or untrusted input. Use XML tags to anchor attention, quarantine injection, and make output parseable.
- [Context Hygiene: Mastering /clear and /compact in Claude Code](https://yaseenkhatib.streamerosai.com/blog/claude-code-context-hygiene-clear-compact/): Claude re-sends the whole conversation every turn, so a messy context window is a recurring bill. How /clear and /compact cut token spend by ~85%.
- [Redux Toolkit Architectures for Massive Enterprise State](https://yaseenkhatib.streamerosai.com/blog/redux-toolkit-architectures/): RTK made Redux's boilerplate reputation obsolete. Structuring feature slices, normalizing with entity adapters, offloading server state to RTK Query — and when not to reach for it.
- [Structured JSON Outputs from Claude with Type-Safe Zod Contracts](https://yaseenkhatib.streamerosai.com/blog/json-structured-outputs-type-safe-zod/): Parsing Claude's prose with regex breaks pipelines. Force type-safe JSON with tool use, then validate it at the boundary with a Zod contract that doubles as your type.
- [The AI-Native Portfolio: Landing Lead Roles by Shipping the System](https://yaseenkhatib.streamerosai.com/blog/ai-native-portfolio-landing-lead-roles/): A portfolio that lists 'used ChatGPT' reads junior. One that demonstrates systems architecture — RAG, agents, guardrails, latency budgets — reads like the lead they're trying to hire. Here's how to build the second one.
- [How I Stop RAG Hallucination With a Grounding Contract](https://yaseenkhatib.streamerosai.com/blog/zero-hallucination-rag-grounding-contract/): Retrieval never stopped hallucination on my streamerOS agent — a grounding contract did: a retrieval score floor, a strict prompt, and required attribution.
- [TypeScript Migration Without a Feature Freeze](https://yaseenkhatib.streamerosai.com/blog/typescript-migration-playbook/): My TypeScript migrations skipped the feature freeze: allowJs coexistence, strictness ratcheted one flag at a time, and any quarantined behind a single typed boundary.
- [Tool Use & Function Calling Mechanics in Claude](https://yaseenkhatib.streamerosai.com/blog/tool-use-function-calling-mechanics/): Claude hallucinates data it should fetch. Tool use fixes it: the model requests a function, your runtime executes it, and the result flows back. The full agent loop, explained.
- [Why I Run RAG at the Edge on Cloudflare Workers + Hono](https://yaseenkhatib.streamerosai.com/blog/edge-native-rag-cloudflare-workers-hono/): The support agent I built for streamerOS runs its whole RAG pipeline at the edge on Cloudflare Workers and Hono, scaling to zero between requests.
- [React useMemo: Measured Wins, Not Cargo-Cult Habit](https://yaseenkhatib.streamerosai.com/blog/react-usememo-benchmarks/): In React reviews, useMemo is the hook I see cargo-culted most: what it truly costs, the two cases that justify it, and why I open the Profiler before memoizing.
- [Model Context Protocol: MCP Server Foundations](https://yaseenkhatib.streamerosai.com/blog/model-context-protocol-mcp-server-foundations/): Hard-wiring integrations into every agent doesn't scale. MCP is a USB-C port for tools: author a server once, and any client discovers and calls it over one protocol.
- [Stateful AI Memory: The Context Window Is RAM, Not Disk](https://yaseenkhatib.streamerosai.com/blog/memory-and-stateful-ai-architecture/): Sable remembers users across sessions, so I tier its stateful memory into a working buffer, an episodic vector store, and durable semantic facts.

## Full archives

- [Full article corpus, plain text](https://yaseenkhatib.streamerosai.com/llms-full.txt)
- [Blog index](https://yaseenkhatib.streamerosai.com/blog/)
- [RSS feed](https://yaseenkhatib.streamerosai.com/rss.xml)
- [Sitemap](https://yaseenkhatib.streamerosai.com/sitemap.xml)