Portfolio · 23 Projects
Building tools
for research & work.
AI tools, research technology, and web applications — designed to remove the friction between an idea and the thing that helps people.
23Projects
shipped
shipped
–Updated
this week
this week
23Live &
deployed
deployed
🧵#17
Research Home — Throughline
Your research home: 22 working papers, Throughline Studio with Research Kitchen, four research libraries and the complete tool directory. Continue an archived paper in Studio with its question and source drafts; reach analysis, journal selection and publication planning from one place.
↗#39
Solo Publish
A dedicated workspace for single authors pursuing genuine publication: six detailed study routes, source-checked journal evidence, a preparation planner, citation and policy ledger, submission milestones, a live journal reality check (real Crossref review/production intervals, recent solo-authored articles, DOAJ fees and review model) with a pool of journals that actually publish single authors, and connections to the existing research tools. Local project storage and exports. Planning assumptions stay separate from verified facts; no acceptance guarantees. Private Site access.
🧵#23
Throughline Studio
A connected A→Z research workspace with fluid stage transitions, distraction-free Focus mode, a unified dark/light interface and a themed research Kitchen. The Research Kitchen guides the journey from pantry to publication: 20 recipes, 12 real R workflows, qualitative coding and mixed-methods integration. Continue through manuscript versions, evidence-based handoff checks, structured reviewer responses and a complete study ZIP with file-integrity hashes. Sources, plans, analysis and writing remain together across seven author-reviewed stages.
⚓#35
AMET CEII — Website Redesign
A ground-up, world-class redesign of the AMET Centre for Entrepreneurship, Innovation & Incubation (aceii.in), pitched by five B.Sc. AI-ML students to redesign and maintain it in-house. Two live design directions: a holistic version (the incubator for all 18 AMET departments) and a maritime-focused version, cross-linked by a switcher — the proposal covers both. Ten pages each (Home, About, Programs, Portfolio, Events, Team & Mentors, Resources, Journal, a multi-step Apply flow, Contact) over a shared premium no-build design system (navy/teal/brass; Newsreader × Hanken Grotesk × IBM Plex Mono), static-first so it stays cinematic under reduced-motion. One Vercel function powers the Apply/Contact intake (emails aceii@ametuniv.ac.in via Resend). Real assets pulled from the live site (official crest, real photos of the CEII team + university leadership, verbatim vision/mission); everything else clearly-labelled sample — no fabrication. Ships with a Registrar-ready proposal doc + pitch deck.
🎓#34
CampusDesk
A full LMS — an authoring portal plus a student exam app — reverse-engineered screen-by-screen from the FACE Prep Authoring Portal. Admins manage institutions → batches → users, author a question bank (MCQ, coding with per-language starter code and Piston-run test cases, reading comprehension), assemble assessments into sections, wrap them in configured assessments (shuffling, tab-switch limits, 6-digit protector codes, result publishing, retakes), and publish to institutions, batches or individual students — with TSV bulk upload, image storage and CSV log exports throughout. Students sign in with their roll number (forced password change), take proctored timed attempts — tab switches and reloads are counted and can hard-submit — run code against visible test cases, and get results per the publishing rules. An AI copilot (Groq) drafts questions and course outlines for the author to review before anything enters the bank, reads cohort stats into instructor insights, and gives students a per-question "explain my mistake" plus a coach debrief — every AI output badged as a draft, never auto-published. An Attendance module enrolls each student from one reference photo, then face-matches a single whole-class photo against the roster — entirely client-side (face-api.js, no photo ever leaves the browser) — to mark present/absent with an editable, overridable results table and per-batch history. Express + Upstash Redis + React 19; seeded with the real 162-user AMET roster.
↻#30
Rethink with AI
Hand a thought to AI and see it differently. Paste anything — a worry, an opinion, a decision — or arrive straight from My Thoughts, then pick a lens: Reframe it (spot unhelpful thinking patterns, find a kinder balanced view), Challenge it (a fair devil's advocate — the strongest counter-case, hidden assumptions, blind spots), Other perspectives (a mentor, a skeptic, a pragmatist, future-you), or Sharpen & act (distill what you mean and get concrete next steps). Switch lenses with no re-typing, or run all four. The AI is framed as a thinking partner, not a therapist — it never diagnoses, invents no facts about you, and every reflection is badged as a prompt to think with, not advice. Static HTML + a Groq-backed Vercel function; your thought is sent only when you ask.
✎#29
My Thoughts
A calm, private place to write and save anything — a quick note, a worry, an idea, a plan. Each thought has a title, free-form body, and tags; everything is full-text searchable, filterable by tag, pinnable, and autosaved as you type. Everything stays on your device (localStorage) — no account, no backend, no analytics — with one-tap JSON export/import to move between devices yourself. Installable as an offline app. A single ↗ Rethink with AI button hands the open thought to the companion app (#30) to reframe, challenge, or sharpen it — and that's the only time any text leaves the device. Single-file HTML, vanilla JS, no build.
📄#31
Callback
A resume & LinkedIn grader for students and early-career applicants. Paste your resume (or load a
.docx), optionally paste a job description, and get a transparent, rule-based scorecard — action verbs vs weak openers (“responsible for…”), quantified impact, clichés, length, and how many of the JD's key terms your resume actually mentions. Every score is a stated heuristic, not a real ATS, and the whole grade runs in your browser — your resume is never uploaded. An optional AI layer sharpens a single bullet using only the facts already in it (missing numbers become a [add metric] placeholder, never a fabricated one). Single-file HTML + one Groq function. Your resume grows as a plant while you paste: one leaf per bullet, filled when it carries a number, with a dot for strong verbs or weak openers.🎙#32
Greenroom
The room before you go on. A mock-interview coach for placement students — pick a track (HR/behavioral, technical viva, aptitude/fit), get a real question, and answer it out loud (on-device speech) or by typing. Greenroom gives an instant, rule-based read of how the answer lands: STAR-structure detection for behavioral answers, filler words, hedging, specificity (numbers/scale) and spoken length — explicitly a rehearsal aid, not a verdict. An optional AI coach reads only your answer and returns strengths, improvements, and one generic STAR skeleton with
[placeholders] — it never invents facts, numbers, or experiences about you. The sibling to Callback: together they cover the paper and the room. Single-file HTML + one Groq function. A practice garden on the setup screen keeps one flower per finished interview, with the five competency scores as petals, stored only in your browser.✂️#36
cutroom
The cutting room — a local studio that turns a described idea into a finished narrated video. Type what you want (explainer · promo · lecture · vertical short) and cutroom writes the script scene by scene, speaks it with a free neural voiceover (7 voices incl. Indian English), finds per-scene HD stock footage (Pexels; animated kinetic-text scenes as the no-key fallback), burns styled pop-in captions timed to the voice, adds title/end cards, fades and ducked music, and renders the MP4 with ffmpeg on your machine. Every scene's narration and visuals are editable on a filmstrip script view before producing — or hit “Roll” and go prompt→picture in one take. A second reel cuts videos from your own footage (upload or YouTube pull: footage / audio / style-only), with AI edit plans sanitized against real clip durations. The interface is a darkened editing suite: film slate, call sheets, live timecode, render queue, screening monitor. No-fab: scripts are barred from inventing stats or citations; edit plans can't reference footage that doesn't exist.
🧭#37
Helm
A from-scratch redesign of FACE Prep's own internal Authoring Portal (admin.faceprep.in) — commissioned from a live batch-management link shared by the Campus Manager. The original drops you straight into a flat data-grid with no home screen; Helm opens on a real command-center dashboard (rollup stats, a "needs attention" feed, score distribution) and replaces the original's "VIEW SUBGROUPS"/"VIEW STUDENTS" modal popups with a genuine hierarchical Institutions → Batches → Students tree explorer. Assessment delivery settings (shuffle · proctoring · results · retake) live inline per assessment instead of a separate collection to hunt down; a Scheduled/Live/Completed publish board replaces the flat administrations list; attempts get a score-distribution histogram + violations panel above the paginated raw log. A ⌘K command palette jumps to any of the 11 modules. Grounded in the real product's schema (reverse-engineered from 56 real screenshots, shared earlier for CampusDesk) — every institution/batch/student/question/attempt inside is simulated demo data, clearly labelled, no connection to FACE Prep's real system.
⏱#28
Journal Timelines
Choose a journal by index coverage and estimated publication speed. Search ~46k Scopus-indexed sources, filter by field, index (Scopus / Web of Science / ABDC), ABDC rating and evidence, then rank by fastest submission → acceptance → publication. The speed estimates come only from real article lifecycle dates — PubMed plus publisher-deposited Crossref received/accepted/online dates (Springer, Wiley, T&F), with each journal's single-author share — journals that don't deposit received/accepted dates show Unknown / needs evidence rather than a fabricated number. Shareable filtered links and CSV export. Static HTML, no build; data built by a no-AI Python pipeline over the official Elsevier Scopus list and NCBI E-utilities.
🪞#26
MirrorScope
Interviews are a mirror of reaction. Paste an interview transcript and decode it per speaker as a sectioned visual report: an at-a-glance scorecard, a turn-taking river and dual-line sentiment timeline, a word cloud, a two-sided word-shift chart, a pronoun ternary and stance map, a reaction map of how they answer each topic, a Q→A scatter, and an annotated transcript with stance words coloured — plus reaction-to-triggers (type a word, see how they responded when the other person raised it). It also measures mirroring — Language Style Matching (Ireland & Pennebaker 2010): how closely the two speakers fall into the same function-word style, with a convergence sparkline — profiles the questioner (open vs closed questions, follow-up and leading-phrasing rates), and reads readability & richness (Flesch, vocabulary, sentence rhythm). Every chart is hand-rolled and computed in the browser (offline, never sent anywhere); an optional AI layer adds a qualitative Big-Five radar + personality read, always badged inference, not a validated score. A passcode-keyed decode library (cross-device, or on-device offline) saves decodes under a person's name — search/filter them, merge variant names, run a combined deep read that pools all of one person's interviews into a single profile, and compare two people side by side. Compare two interviews to see the stance drift between them. Single-file HTML + four Vercel functions. The pasted transcript appears live as reeds on a waterline: one reed per turn, height = words, main speaker above and the other reflected below.
🪣#33
Runway
Days of money left. A private daily-expense & EMI recovery dashboard: say "spent ₹30 on milk" — typed or by voice — and an offline parser (English/Tamil/Urdu/Hinglish keyword map, AI fallback for messy sentences) records and categorises it instantly. The home screen is a water-level bucket for the month and a coin stack for today's allowance — a coin drops every time you spend. An EMI lane tracks every loan's due day with paid/unpaid ticks, months left, the honest EMI-road-left total, and the projected debt-free month; the month view shows spend-by-category, day-by-day pace against the budget line, inflows, and the net position — negative is the point: the recovery number to shrink. Passcode-keyed online book (no account, cross-device), calendar-month budgets, backup export. Single-file HTML + three Vercel functions.
{ }#24
Throughline CS
The Throughline idea, rebuilt for computer-science researchers — from problem to paper in seven stages. One search across arXiv · Semantic Scholar · DBLP · OpenAlex (live, deduped, source-badged) builds a reading list that becomes a real BibTeX; a novelty check runs your research questions against the literature; an experiment matrix seeds a run tracker; verified statistics (paired t / Welch, Cohen's d, t-based CIs) aggregate runs across seeds into capture-ready Markdown/LaTeX tables; a hand-verified venue directory (DBLP-linked) closes the loop; and a Top-venue desk drafts an area-chair-audited paper plan in eight stages, grounded in live arXiv, each open to a NeurIPS-style Reviewer 2. Static ES modules, no build, local-first, nothing fabricated. The home page opens on a river of computer science: 11.85 M OpenAlex works across 11 CS subfields, 2000–2025, live.
🔎#22
BookScope
A curated library of 284 academic reference books — roughly 128,000 source pages of research-methods, statistics, and social-science texts. Search available extracted text in the static Pagefind index, browse titles and authors, then open the source. Some books lack searchable text, and very long extracts are capped. Each book now opens an “Inside this book” panel — constructs and theories with page numbers, distinctive terms, its own table of contents and related books — plus a library-wide concept atlas. Search runs entirely client-side; metadata comes from real source files or provenance-labelled OpenLibrary matches. The book-length companion to Syed's Research Book. The hero is a grove: one tree per book, placed by year, trunk height = page count, colour = topic, and your search lights up the matching books.
📚#21
Syed's Research Book
A browsable, time-flowed view of a 24-month systematic-review corpus — 9,388 hand-coded papers across 167 management / organizational-behavior constructs. Alongside 79,517 recent OpenAlex records, a separate all-years management import adds 1,863,750 real records, 1,165,306 provider abstracts, and 58,487,941 observed reference links. All enabled tiers total 1,952,655 records. The official OpenAlex 2026-09-23 release traversal and Crossref source traversal are complete for the recorded selection; missing titles and unresolved reference identities remain explicit. This is not universal literature or citation coverage. Source selection uses 2025 highest-category CiteScore rankings. Conference series, book series, trade publications, and journals stay separate. Explore the 408-construct co-occurrence map, construct trends and Term Trends. Hand coding stays intact; imported tags, extracted phrases and catalog crosswalks are machine-matched and verify-badged. The full snapshot is hosted on Vercel; the older GitHub Pages version remains available separately.
🔬#20
Paper Dissection Workshop
↗ Folded into PaperCards (#16) — one "Paper analysis" tool instead of two. Dissect a paper to the core: theory, design, sample, measures, analysis, software, findings and more, on a facet × year timeline + synthesis matrix, extracted only from the source text and grounded with verbatim evidence — nothing invented. Now lives as the "Dissection" depth inside PaperCards.
⚠️#19
FallacyScope
A curated field guide to the paradoxes & fallacies that bend social-science & management research — Simpson's, ecological, base-rate, survivorship and more. Each entry: the trap, a real worked example, why it happens, how to avoid it, and where it bites in the research lifecycle. Every one hand-cited to a canonical source with a DOI link. Now 87 entries (19 paradoxes · 27 fallacies · 41 biases); all are now embedded in ResearchFlow as a stage-aware resource. The front page is now a stream: one stone per trap at the research stage where it first appears, trailing to the last stage it affects.
🎓#18
ScholarScope
Who leads — and where to publish — in social science & management. Live OpenAlex discovery and a Publish desk, plus the new Author Explorer (construct-level dossiers for 21,961 authors: fingerprint, focus over time, construct pairings, collaborators, and every publication newest-first — corpus papers plus the full OpenAlex career) and Journal Explorer: visual journal trajectories, topic evidence, verified editorial opportunities and a backward submission planner. Journal of Organizational Behavior is the first detailed editorial dossier. Topic outlooks are labelled inferences, never acceptance probabilities. The front page opens on a live mountain range: the top 26 institutions in each lens field as peaks, height = their works in that field (OpenAlex), each peak one click from its profile.
🔬#16
PaperCards
Research papers as glanceable cards — drop a DOI or abstract and AI extracts the constructs, effects, and methodology into a construct graph, forest plot, and method pills. Or dissect a paper to depth: 13 evidence-grounded facets (theory, design, measures, analysis…), a facet × year timeline, and a synthesis matrix. Compare papers and explore their citation networks. (Now includes the former Paper Dissection Workshop.)
📊#15
ToolsScope
The Research Suite's analysis bench: twelve executable R workflows including lavaan, lme4, mice, metafor, fixest, survival, mgcv and quantreg, alongside core statistics and qualitative coding. Twenty-four verified method sources, actual-output plots and reproduction ZIPs; the same engines now run directly inside Studio’s Research Kitchen.
🏛️#14
TheoryScope
A curated library of 449 social-science & management theories across 24 disciplines, from Taylor (1911) to generative-AI augmentation. Browse, AI-match to your research question, compare famous debates, and explore the influence and construct maps — or drop research papers in by DOI onto a shared Canvas, rendered in the same flow template as the theories. Each card cites its sources, links to its scales in ScaleScope, and exports APA, BibTeX, and RIS.
🧭#13
Tracewise
Map the work you actually do — the real as-is process, not the SOP. A hybrid interview traces each step and tags its friction; the Automation-Fit engine then scores where AI belongs versus what to protect. The Ecosystem view stitches roles together and stress-tests what breaks first under 10–15× growth.
🔬#08
ResearchFlow
AI-powered 10-stage wizard that walks researchers from raw idea to publication-ready methodology — question, theory, hypotheses, design, and full analysis plan.
🗺️#03
KarmaMap
Interactive D3 treemap visualising AI's impact on Indian job roles. Claude AI scores each role by growth rate, salary, exposure level, and future outlook.
🌐#04
WordMap
↗ Folded into the Library — Wordmap's animated 50-year term treemap (425 OpenAlex social-science terms, 1970–2054) is now the Term Trends view inside Syed's Research Book. The standalone URL redirects there.
✅#01
TaskFlow
Full-stack team task manager with JWT auth, role-based access for managers and mentors, and multi-assignee task workflows. Deployed on Vercel with Upstash Redis storage.
📚DOCS
Documentation Master
SOP ownership desk with the 17-person campus roster: accountable owners, backups, step assignments, completion evidence and versioned approval reviews. Answers appear in the procedure reader, handbook and Word/PDF exports. A private evidence workspace for cases, quoted findings, gap review, SOPs and leadership reports. Source-backed claims retain their evidence.
↗HOME
Launchpad — Project Home
All projects in one place, organized into eight collections: research, libraries, campus, work, careers, personal, creative and merged projects. Search, frequently opened apps, HTTP status checks and clear local/private labels.
📖SOP
Quiet Operations
A FACE Prep campus SOP library: 12 practical operating procedures with document control, responsible roles, numbered actions, expected results, exceptions and completion evidence. Individual two-page PDFs, mobile instruction cards and a separate automation appendix. Operator edition v1.1.0; approval and effective dates remain pending. Original design volumes preserved.
🌱#38
Common Ground
A playable 3D campus for FACE Prep operations. Walk with keyboard or touch controls, visit buildings, meet mentor characters and discover six places. Real quests, a planner, teaching schedules and a growing contribution garden connect play to daily work. A manager-only work archive holds numbered summaries and searchable source history. Linked month, week and day views organize priorities around second- and fourth-Saturday working weeks. Existing campus accounts protect real records. No combat, rankings or streak penalties.
🎓#25
FACE Prep Campus
Campus operations home for FACE Prep @ AMET University. The admin Daily Work desk connects retained WhatsApp/email requests and stakeholder follow-ups to owned mentor tasks, due dates, waiting checks and evidence-based closure. One shared task record, explicit feed freshness and a downloadable work brief. Separate admin / mentor / student interfaces. Per-batch & mentor-wise timetables (B&W PDF + substitution planner), per-session attendance with class-incharge leave, fees, placement drives with a conversion-funnel tracker, student-readiness DB & a 90%-target outcomes dashboard (now a convergence hub: CampusDesk assessment scores, Callback resume grades and Greenroom mock-interview verdicts flow into one blended per-student readiness view — students opt in per result, missing signals flagged, never estimated; plus one-click roster sync from the org's private Google tracker sheets via a service account), per-subject materials, secure online assessments (master bank → random per-student subset, fullscreen copy-lock), learning workshops, staff HR & performance, holiday-aware academic calendar, a daily Classroom Infrastructure tracker (AC/WiFi/plug points, whichever of projector or smart board a room has, plus each room's student strength — flagged issues + CSV/Excel/PDF exports for AMET management), and a Command Center that mirrors the campus WhatsApp groups (mentor/student/parent messages, auto-tagged by phone) via a self-hosted bridge, plus an Email Center that unifies the campus email IDs into one admin inbox — monitor, reply in-thread & broadcast to a class or role, with delivery tracking and keyword alerts. Personal day/week/month/semester planning with magazine-style reading, PDF downloads, phone-calendar alerts and recorded outcomes; Android companion preview with native schedule reminders, Now/Next notification card, connected tasks/follow-ups/SOP ownership queue, and widget. JWT auth, bulk CSV/Excel upload, Upstash Redis. Installable as a mobile app (PWA) — an optional “add to home screen” prompt on phones, full-screen & offline-capable, with mobile-friendly safe-area & tap polish. Deployed on Vercel.
📰#05
JournalTime
Research writing workspace for social-science academics. Research Compass (AI gap finder), Article Developer (AI full-draft), Journal Studio — a guided "write for this journal" flow — and Journal Formatter: upload a manuscript, download it in a journal's submission format, references restyled from real Crossref records only. Finding the right journal lives in ScholarScope.
📐#09
ScaleScope
Library of 330+ validated measurement scales across 7 domains — browse, get AI recommendations, and compare scales with APA 7th citations.
🧭#12
Career Compass
Evidence-based 8-stage career counselling instrument for Class 10 students. 115+ items covering RIASEC, Big Five, aptitudes, and values — ends in a personalised stream report.
💡#11
IdeaBox
Mobile-first idea capture for thoughts on the go — voice or text, auto-categorised, tagged, pinned. A standalone capture tool (unbundled from the research suite in the simplification); it still accepts a
?topic= hand-off that pre-fills the textarea.🎵#10
Cadence
Immersive survey tool for longitudinal research. One-question-at-a-time flow, milestone celebrations, keyboard shortcuts, and temporal wave tracking with participant codes.
📅#07
Timetable Generator
Constraint-based academic timetable generator for 11 classes across 3 departments. Respects mentor availability, subject caps, and AY 2026–27 scheduling rules.
🍳#06
Bachelor Meal Plan
"What's in my kitchen?" — type your ingredients and it matches 200+ halal recipes offline ("Cook now" / "Almost there"), or asks AI to invent a fresh one. Plus weekly planning, shopping lists, and portion guidance. Mobile-first, localStorage, no sign-up.
🧬#02
Nexus
The portfolio you're looking at right now. Single-file HTML with light/dark theming, a story-grouped project grid (suite · campus · personal), fuzzy search, clickable tag facets, live GitHub commit sparklines, honest deploy-reachability dots, a push-date timeline view, and per-card share links — a living index of everything built.
Current view
No projects match
Try a different search or .
Artifacts
Living documents & one-off tools that get updated over time
📚
IO Psychology Research Tracker
26 papers from Journal of Applied Psychology, Personnel Psychology, and J. Organizational Behavior (Feb–May 2026). Filter by journal, search by keyword, and expand for Claude AI–generated takeaways.