Skip to content
View milanpatel98's full-sized avatar

Block or report milanpatel98

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
milanpatel98/README.md

Hi, I'm Milan 👋

Business Analyst. I learn how a business really runs, then turn it into clear reports, simpler processes and, when it helps, the software that fixes the problem.

I've worked inside hotel operations, reporting with SAP and Power BI. I hold an M.S. in Computer Science and I'm pursuing a Doctor of Business Administration at Westcliff University. I also design and build hotel software for The Neural Horn.

Portfolio: milan.us · CV: milan.us/cv · LinkedIn: milanpatel72 · X: @milanpatel72 · Email: me@milan.us


Projects with measured results

Every project here has a live demo and a case study. Each one says what was tested, on what data, and what isn't proven yet.

Project What it does Result Links
Circa Ask a patient's chart a question. Every sentence of the answer cites its source record. 97.8% correct on 1,681 held-out questions, vs 66.1% for search-based retrieval demo · case
Countersign Real-time card fraud scoring that explains every decline. 88% of fraud caught reviewing 0.5% of payments, vs 44% for rules demo · case
4th Live NBA scores with each team's chance to win, every play. Calibrated on a season it never saw; 23 games a wrong call may have decided live · case
Driftless Raft consensus in Go, checked by a simulator that tries to break it. 2,000 runs, 43,410 injected faults, zero broken invariants; every planted bug caught demo · case
Mender A code-repair agent with a benchmark that catches cheating. Real fixes solve 13/13; a scripted cheater scores 0/13 demo · case
Hark Live captions on your device that learn your voice from your corrections. Misses on corrected names 18/22 → 13/22, accuracy never worse demo · case

Products (code private): Caelis, ID scanning that fills the hotel booking form in about 3 seconds · Takeover, a 124-item playbook and toolset for hotel ownership changes · MindBridge AI, research on early signs of depression and anxiety in everyday conversation.


What I work with

  • Business analysis: requirements gathering, process mapping (BPMN, Visio), gap analysis, BRD / FRD, requirements traceability, Agile / Scrum, JIRA & Confluence
  • Data & BI: SAP, Power BI (KPI & dashboard design), Tableau, SQL, Excel, Python (pandas, NumPy), statistics & model evaluation, FHIR healthcare data
  • Software: TypeScript, React & Next.js, Node.js, Go, Swift, Supabase, Cloudflare, AWS
  • AI/ML: PyTorch, scikit-learn, fine-tuned LLMs, NLP, RAG & evals, on-device speech recognition

Open to Business Analyst roles. If your team needs someone who can map a messy process and make it measurable, say hi: me@milan.us

Pinned Loading

  1. 4th 4th Public

    Live NBA scores with win chance, shot charts and the calls the refs got wrong

    JavaScript

  2. circa circa Public

    Ask a patient's FHIR record a question; every sentence of the answer cites its source record and is checked. Held-out benchmark vs RAG. Live demo in the browser.

    Python

  3. countersign countersign Public

    Real-time card fraud scoring that explains every decline, with features that cannot drift between training and serving. Live demo in the browser.

    Python

  4. driftless driftless Public

    Raft consensus in Go, checked by a deterministic simulator: seeded partitions and crashes, a hand-written linearizability checker, and planted bugs it must catch.

    Go

  5. hark hark Public

    Live captions that run on your device and learn your voice from your corrections.

    JavaScript

  6. mender mender Public

    A code-repair agent built from first principles, with a benchmark that catches cheating: hidden tests, tamper detection, prompt-injection traps.

    Python