jaylin man

v1.0 · oct 2026

Jaylin Man

I talk to the people doing the work, then I build the thing that saves them time. Lately, a lot of that is AI tooling: for developers, and for the agents they work with.

now
Technical Product Intern at Stutter Forward, Product Engineer at Asian American Dream, and building Schedulr for a 40-person restaurant
before
AI Research Intern, Air Force Research Laboratory
won
1st place at the Cisco AI Case Competition
studying
Computer Science at Binghamton University, class of 2028
open to
Summer 2027 internships in product management, technical PM, product engineering, tech consulting, solutions engineering and AI engineering

changelog

Experience

Newest first. Most started by talking to the people doing the work.

  1. - now · binghamton, ny

    Technical Product Intern @ Stutter Forward

    Leading development of an online stuttering-support platform for a clinical program with a 72-person waitlist.

    • Designing an automated sign-up flow to replace a 5-step manual enrollment that takes 2 days per cohort.
    • Turning program needs into a phased roadmap with 2 clinicians, targeting a working prototype within 3 months.

    product discovery · roadmapping · healthcare

  2. - now · binghamton, ny · independent

    Technical Consultant & Developer @ Schedulr

    Cut weekly scheduling for a 40-person restaurant from 2 hours to 15 minutes by replacing WhatsApp threads and spreadsheets.

    • Shipped shift swaps, surveys and PDF exports across 6 live releases, prioritized from interviews with 5 staff members.
    • Built a greedy constraint-based scheduler that ranks staff by weighted scores and proposes 5 schedule options.
    • Set up a docs harness where coding agents follow repo rules to write release notes and update docs with every change.

    client work · user interviews · scheduling algorithms · agentic dev

  3. - · rome, ny

    AI Research Intern @ Air Force Research Laboratory

    Produced an ontology of 1,103 institutions for an institutional-trust study with a Python LLM and embeddings pipeline.

    • Built a CLI agentic memory system on MCP that gives coding agents architecture maps instead of raw source files.
    • Let analysts score and chart text data without writing code, through a drag-and-drop AI workflow studio with 78 reusable operators.

    python · llms · embeddings · mcp · agents

  4. - now · new york, ny

    Product Engineer @ Asian American Dream

    Processed 400+ RSVPs across 3 events by launching Luma-style event pages with QR check-in and automatic reminders.

    • Added SMS alerts to the mentorship portal with AWS SNS for events, RSVPs and reminders, alongside its SES email.
    • Redesigned the v2 dashboard, profiles and mobile layout, prioritized from interviews with 10 mentors and mentees.

    product engineering · user interviews · aws sns + ses

  5. - · queens, ny

    Software Engineer Intern @ Health Guard Pharmacy

    Cut payroll prep from 3 hours to 10 minutes by replacing Excel sheets at a 30-employee pharmacy with a clock-in web app.

    • Gathered requirements from lead pharmacists to fit real shifts, time-off requests and schedule changes.
    • Designed GPS-verified clock-ins within a 100-meter radius, tied to scheduled shifts, so managers could trust every timesheet.

    requirements · web app · geolocation

Skills

languages
Python, TypeScript, JavaScript, Java, C, C++, SQL
building
React, Next.js, Node.js, FastAPI, REST APIs, Postgres, Supabase, Firebase
cloud
GCP (Vertex AI), AWS (SES, SNS), Docker, Vercel, Cloudflare Workers, Git
ai
MCP, Claude Code, Codex, Opencode, ChromaDB, embeddings
product
user interviews, roadmaps, GA4, Jira, Excel, PowerPoint

repos

Projects

Cool tools I built. The public ones link to their code.

  • Gives coding agents a map of a codebase instead of the raw source.

    docpipe generate , then .context-tree/ (3 tiers) , then docpipe mcp , then your coding agent

    Agents read a one-page repo map first and pull only the file contracts they need, through four MCP tools. Unchanged files are hashed and skipped, so re-indexing stays cheap.

    from 31.4M to 1.88M tokensraw repo vs. whole index, one 1M-line Java repo

    python · mcp · llm pipelines · cli

  • afterword

    active

    Live meeting transcripts that know who's talking.

    system audio + mic , then Gemini, live , then voiceprints , then named transcript , then action items by owner

    A wrong name is worse than no name: a name sticks only when several independent answers agree and its quote is really in the transcript. Everyone else stays Speaker A, B, C.

    from 67.7% to 92.6%speech credited to the right person, one real 37-minute call

    python · gemini · speaker recognition · pyside6

    code private for now · walkthrough on request

  • kira

    prototype

    A personal AI agent that asks before anything risky.

    tool call , then risk level × autonomy mode , then allow / ask / block , then approve on Telegram , then audit log

    73 tools across mail, files, shell and git, each with a declared risk level. The gate runs in plain Python outside the model, so the model never grants its own permissions. Memory lives in pgvector.

    python · agents · vertex ai · postgres + pgvector · telegram

    code private for now · walkthrough on request

  • Every git branch, PR and conflict in one command, for you and your agents.

    vb , then branch tree + PR and CI state , then predicted conflicts , then needs you, with fix commands

    One map, two audiences: a tree, TUI or local web view for people, and compact text, JSON or a read-only MCP server for agents.

    from 1,524 to 417 tokensfor an agent to read the branches of a demo repo, 11 git calls down to 1

    go · git · mcp · devtools

off the main branch

Beyond work

  • 1st place · Cisco AI Case Competition

    -

    Co-built 2 of 3 working prototypes, including a meeting-briefing assistant, then pitched a phased rollout to 4 Cisco executives through live skits set in a real workday.

  • VP of Internal Affairs · StackHacks

    - now

    Automated weekly feedback loops for 50% fewer meetings, moved the board and members onto one Discord, and led hackathon marketing for +25% engagement.

  • Peer Mentor · CodePath

    Mentored 7 students.

  • Hacker · HackHarvard 2025

  • IT Chair & Historian · Hong Kong Exchange Square

Education

B.S. Computer Science · Binghamton University (SUNY)

gpa 3.71 · dean's list, fall 2024 - spring 2026

Off the clock

  • NYC-born, Long Island-raised, upstate for school.
  • Will challenge you to pickleball. Also spikeball.
  • Lifts, boulders and swims regularly.
  • Plays piano and likes to make short films.
  • Designs graphics for student orgs and makes content on the side.

roadmap

What's next

now
Building Stutter Forward's first prototype with two clinicians, shipping Schedulr releases, and studying CS at Binghamton.
next
Summer 2027 internships in product management, technical PM, product engineering, solutions engineering or AI engineering.

Email is the fastest way to reach me.