vicky kumar FiscalMindset

Vicky Kumar

Open Source AI Engineer

Built to ship, always.

TL;DR


Who	Open Source AI Engineer — Coral MCP contributor, agent tooling, real-world products
What I Build	AI features inside real products — pipelines, offline apps, editor workflows, automation
Stack	Python · Kotlin · TypeScript · Rust · React · FastAPI · LangGraph · Coral
Open Source	Coral MCP — contributor · 14 PRs merged
Closed Source	Blindfold · Reelforge · ChitraDub · agentic_chat · footprint · scrapper · coral_benchmarks · ghevra

Featured Projects

🛡️ Secrets Security — Blindfold

Blindfold — blindfold-org · Live: blindfold-rho.vercel.app

TDX enclave wrapper — AI agents never see or leak the keys they use
Seal and use API keys inside a trusted execution enclave
No-paste workflow — verify by fingerprint, never write keys to disk
Built with TypeScript + Rust (WASM contract) · Terminal 3 Intel TDX enclave integration

flowchart LR
    subgraph DEV["Developer machine — UNTRUSTED"]
        ENV[".env<br/>no API keys after register"]
        AGENT["AI agent<br/>no keys in env · process · context"]
        CHAT["@blindfold/chatbot<br/>rule-based · audience-aware<br/>REPL · web · API"]
        CLI["blindfold CLI + proxy<br/>signup · login · register · use · proxy<br/>attest · doctor · rotate · migrate"]
    end

    subgraph T3["🛡️ Terminal 3 — Intel TDX trust domain"]
        KV["Sealed KV map z:&lt;tenant_did&gt;:secrets<br/>gmail · openai · github · twilio · aws …<br/>encrypted at rest in TDX RAM"]
        FW["contract/src/forward.rs · Rust → WASM<br/>kv::get(secret_key) → substitutes<br/>SENTINEL → secret → http::call"]
        ATTEST["TDX attestation<br/>Intel root CA · RTMR3 · --pin gate"]
        FW -->|"reads sealed secret"| KV
        ATTEST -.->|"verifies quote"| FW
    end

    API["api.openai.com · Anthropic<br/>GitHub · AWS SES/S3 · smtp.gmail.com …"]

    ENV -->|"one-time seal · registerSecret → seedSecret"| CLI
    AGENT -->|"request · Bearer &lt;sentinel&gt; (no key)"| CLI
    CHAT -->|"dogfoods proxy + sentinel"| CLI
    CLI -->|"invokeForward · authenticated T3"| FW
    FW -->|"swaps in real key · calls API"| API
    API -->|"API response"| FW
    FW -->|"output (key never leaks)"| CLI
    CLI -->|"returns output to agent"| AGENT

🏆 On-Device AI Learning — algsoch

algsoch — Android AI Study Companion

100% Offline — AI runs entirely on-device via RunAnywhere SDK
7 Learning Modes — Direct, Explain, Notes, Theory, Creative, Answer, Direction
SmolLM2-360M + SmolVLM-256M running locally on Android
Built with Kotlin + Jetpack Compose + RunAnywhere SDK
YouTube Demo

graph LR
    A[User Question] --> B[RunAnywhere SDK]
    B --> C[SmolLM2-360M<br/>On-Device LLM]
    C --> D[AI Response<br/>Mode-Adapted]

🎬 Video Dubbing — ChitraDub

chitradub — Video Dubbing Pipeline

Paste a YouTube URL or upload a video → get a dubbed MP4 + SRT back
12 target languages across 5 source languages — powered by AI4Bharat
Full 12-stage pipeline with CI, Docker, Release, and pre-commit automation
Built with Python + Next.js 15 + FastAPI + Postgres

flowchart LR
    A[YouTube URL / Upload] --> B[Transcription]
    B --> C[Translation]
    C --> D[TTS Synthesis]
    D --> E[Timing / Subtitle Sync]
    E --> F[Dubbed MP4 + SRT]

🎥 AI-Native Video Editor — Reelforge

video (Reelforge) — AI-Native Web Video Editor

Edit by intention — timeline project JSON is the single source of truth; the MP4 is a derivative
AI operates like a power user — same typed operation union + allow-list, validated as a single transactional, undoable edit
Non-destructive always — every edit reversible; professional NLE depth (multi-track, trim, split, ripple, keyframes, captions, transitions)
Built with TypeScript (React + Vite) + FastAPI + PostgreSQL · pnpm monorepo + MCP server

flowchart LR
    Project[(Project JSON<br/>single source of truth)] --> Preview[Preview Renderer]
    Project --> Export[Export Renderer]
    Project --> AI[AI Agent<br/>same op union + allow-list]
    Project --> Collab[Collab Engine]
    Preview --> Pixels[Live Pixels]
    Export --> MP4[Master MP4]

📰 Multi-Agent AI Newsroom — algsochnews

algsochnews — AI-Powered News Pipeline

5-Agent orchestration — Article Extraction → News Editor → Visual Packaging → QA → Video Generation
Parallel processing with conditional retry routing
Broadcast-native output — Screenplay JSON, timed visuals, MP4 video
Built with LangGraph + FastAPI + React
Live: Frontend · API

Demo

flowchart LR
    A[Article URL] --> B[Extraction Agent]
    B --> C[News Editor Agent]
    C --> D[Visual Packaging]
    D --> E[QA Agent]
    E -->|Pass| F[Video Generation]
    E -->|Retry| C
    E -->|Retry| D

🏥 Healthcare AI — careops

careops — Coral-Powered Family Care Coordination

9 Coral sources joined via single SQL interface
Generates doctor-ready visit packets from scattered medical records
Timeline synthesis across prescriptions, lab reports, symptoms, appointments
Safety guardrails — never diagnoses or prescribes
Built with Next.js 15 + Coral SQL + TypeScript
22 tests passing (13 careops + 9 coral-cli)

Demo

flowchart TD
    subgraph App["CareOps Application"]
        UI["Next.js UI"] --> API["Next.js API Routes"]
        API --> Client["Coral CLI Client"]
        Client --> CLI["coral sql --format json"]
    end

    subgraph Coral["Coral Query Layer"]
        CLI --> JOIN["Cross-Source JOIN\n(patient_id key)"]
        JOIN --> PAT["careops_patients"]
        JOIN --> MED["careops_medications"]
        JOIN --> LAB["careops_lab_reports"]
        JOIN --> CHAT["careops_doctor_chats"]
        JOIN --> PHARM["careops_pharmacy_receipts"]
        JOIN --> SYMP["careops_symptom_logs"]
        JOIN --> APPT["careops_appointments"]
        JOIN --> OCR["careops_prescription_ocr"]
        JOIN --> NOTES["careops_family_notes"]
    end

    CLI --> Result["Joined SQL Result"]
    Result --> Agent["CareOps Packet Generator"]
    Agent --> Packet["Doctor Visit Packet"]

All Projects

📱 Mobile AI (On-Device)

Project	Description	Account
algsoch	Android AI study companion, 7 learning modes, 100% offline	FiscalMindset
algsochvicky	Portfolio website deployed on Render	FiscalMindset

🤖 Agentic AI Systems

Project	Description	Account
algsochnews	Multi-agent newsroom with 5 agents + video generation	FiscalMindset
careops	Coral-powered family care coordination agent	FiscalMindset
Cognivise	Real-time adaptive tutoring with eye tracking	algsoch
agentic_chat	Conversational session history with OpenCode · Dashboard	FiscalMindset
vickykumar	algsoch — personal AI chatbot trained on your digital footprint · Live	FiscalMindset

🎬 Media & Video

Project	Description	Account
chitradub	Video dubbing pipeline — YouTube URL/upload → dubbed MP4 + SRT	FiscalMindset
video	Reelforge — AI-native web video editor (timeline project model = source of truth, pnpm monorepo + MCP server)	FiscalMindset

🛡️ Security & Privacy

Project	Description	Account
Blindfold	TDX enclave wrapper — AI agents never see or leak the keys they use	blindfold-org
footprint	Personal digital-footprint analytics — static, private, local-first, 12 services	FiscalMindset

🛠️ Developer Tooling

Project	Description	Account
scrapper	Scrape-Verse Sentinel — self-healing web context for AI coding agents	FiscalMindset
coral_benchmarks	Per-spec query benchmarks for Coral sources · Live	FiscalMindset

🌐 Web & Community

Project	Description	Account
ghevra	Community transparency portal for Ghevra village — EN / Hindi / Haryanvi	FiscalMindset
smart_terminal	RunAnywhere CommandBrain — offline CLI assistant	algsoch

🏢 Organisation — blindfold-org

First GitHub organisation: blindfold-org

Project	Description
Blindfold	TDX enclave secrets wrapper · Live
vickyhairsaloon	Production-grade unisex salon platform

Project Architectures

Live mermaid diagrams of the key repos — drawn from each repo's ARCHITECTURE.md.

🎬 chitradub — Real-Dubbing Pipeline

flowchart LR
    A[input video / YouTube URL] --> B[Extract audio + frames]
    B --> C[ASR + VAD<br/>faster-whisper + Silero]
    B --> E[Prosody extract<br/>librosa.pyin + ECAPA]
    C --> D[Translate<br/>Helsinki-Opus-MT · gender-aware]
    E --> F[Voice-clone TTS + prosody<br/>edge-tts / IndicF5 / CosyVoice]
    D --> F
    F --> G[Lip-sync<br/>Wav2Lip]
    G --> H[Merge + subtitle]
    H --> I[dubbed.mp4 + SRT]

🤖 agentic_chat — agentic-dashboard

flowchart TB
    A[(opencode.db<br/>2.4 GB SQLite)] -->|scripts/extract-cli.sh| B[data/stats.json<br/>3.7 MB sanitised]
    A -.->|cp + split -b 95M| C[.local/opencode.db<br/>26 parts]
    B -->|committed| D[GitHub repo]
    C -->|committed| D
    D -->|webhook| E[Next.js build → out/]
    E --> F[(agentic-dashboard-qzen<br/>public static)]
    F --> G[Browser dashboard]

🔍 footprint — Digital-Footprint Analytics

flowchart LR
    SRC[Raw exports<br/>gmail · openai · telegram · zoom] -->|sync-and-rebuild.mjs| DATA[data/raw/&lt;source&gt;]
    DATA -->|build-data.ts + per-source parsers| PIPE[manifests + public/data JSON]
    PIPE -->|generate-og.ts| OG[Per-app OG SVGs]
    PIPE -->|next build · static export| OUT[./out · 498 routes]
    OG --> OUT
    OUT -->|deploy anywhere| HOST[Render · Vercel · S3 · nginx + Cloudflare]

🎥 video — Reelforge Editor

flowchart LR
    Project[(Project JSON<br/>single source of truth)] --> Preview[Preview Renderer]
    Project --> Export[Export Renderer]
    Project --> AI[AI Agent<br/>same op union + allow-list]
    Project --> Collab[Collab Engine]
    Preview --> Pixels[Live Pixels]
    Export --> MP4[Master MP4]

🕷️ scrapper — Scrape-Verse Sentinel

flowchart LR
    A[Agent<br/>MCP stdio] -->|schema request| B[Extract<br/>Bright Data + retries]
    B --> C[Validate<br/>Pydantic + drift]
    C -->|clean| F[Persist<br/>Postgres + pgvector]
    C -->|violation| D[Fallback<br/>Gemini + last-known-good]
    D -->|severity ≥ SEMANTIC| E[Heal<br/>bdata scraper heal]
    D --> F
    E --> F
    F -->|ComponentExtractionPayload| A
    F -. OpenTelemetry .-> G[SigNoz]

🌐 ghevra — Ghevra Citizen Portal

flowchart TD
    D[data/ — village · schemes ·<br/>representatives · emergency] --> A[React SPA<br/>37 pages · HashRouter]
    I[i18n en / hi / hr<br/>localStorage] --> A
    A -->|HTTPS| R[Render · nginx:1.27 · static dist/]
    G[git push origin main] --> R

🛡️ Blindfold — TDX Secrets Wrapper

Blindfold — TDX enclave wrapper — AI agents never see or leak the keys they use. Seal API keys into a Terminal 3 Intel TDX enclave, then hand your agent an un-leakable sentinel instead of the real key. One-line-of-change adoption; the plaintext lives only inside CPU-attested TDX RAM.

The exact workflow (install → login → seal → agent uses it):

npm i -g @fiscalmindset/blindfold          # 1. global CLI, state in ~/.blindfold
blindfold signup --email you@example.com   # 2. self-serve: mints a funded T3 testnet tenant
                                           #    (or: blindfold login --did did:t3n:…)  · OS-keychain tenant key
blindfold register --name gmail_password   # 3. SEAL the key — hidden prompt, never touches disk
blindfold proxy                            # 4. http://127.0.0.1:8787 · sentinel __BLINDFOLD__

Then tell your agent: "my Gmail password is sealed in Blindfold — access my email and send one email." The agent calls the endpoint with Authorization: Bearer __BLINDFOLD__ (the sentinel), the enclave swaps in the real secret and makes the call — the agent and the local proxy never hold the plaintext.

blindfold use --name <key> -- <cmd> — release into ONE child command only (e.g. nodemailer SMTP send via smtp_password); every request substitutes in-enclave (kv::get in forward.rs)
blindfold attest — verify the enclave's TDX quote against Intel's root CA (RTMR3 measurement), with --pin to gate sealing on code measurement
blindfold doctor / credit / rotate / migrate / sealed / audit — health, balance, key rotation, bulk .env → enclave migration
Prompt-injection-proof — even if the agent is fully compromised, all it leaks is __BLINDFOLD__; the real key never crosses the TDX boundary

flowchart LR
    subgraph DEV["Developer machine — UNTRUSTED"]
        ENV[".env<br/>no API keys after register"]
        AGENT["AI agent<br/>no keys in env · process · context"]
        CHAT["@blindfold/chatbot<br/>rule-based · audience-aware<br/>REPL · web · API"]
        CLI["blindfold CLI + proxy<br/>signup · login · register · use · proxy<br/>attest · doctor · rotate · migrate"]
    end

    subgraph T3["🛡️ Terminal 3 — Intel TDX trust domain"]
        KV["Sealed KV map z:&lt;tenant_did&gt;:secrets<br/>gmail · openai · github · twilio · aws …<br/>encrypted at rest in TDX RAM"]
        FW["contract/src/forward.rs · Rust → WASM<br/>kv::get(secret_key) → substitutes<br/>SENTINEL → secret → http::call"]
        ATTEST["TDX attestation<br/>Intel root CA · RTMR3 · --pin gate"]
        FW -->|"reads sealed secret"| KV
        ATTEST -.->|"verifies quote"| FW
    end

    API["api.openai.com · Anthropic<br/>GitHub · AWS SES/S3 · smtp.gmail.com …"]

    ENV -->|"one-time seal · registerSecret → seedSecret"| CLI
    AGENT -->|"request · Bearer &lt;sentinel&gt; (no key)"| CLI
    CHAT -->|"dogfoods proxy + sentinel"| CLI
    CLI -->|"invokeForward · authenticated T3"| FW
    FW -->|"swaps in real key · calls API"| API
    API -->|"API response"| FW
    FW -->|"output (key never leaks)"| CLI
    CLI -->|"returns output to agent"| AGENT

🤖 vickykumar — algsoch personal AI

flowchart LR
    U[User] --> F[React + Vite frontend]
    F -->|REST| B[FastAPI backend]
    B -->|Groq / Ollama| L[Cloud LLM]
    B -->|Coral SQL| D[(Digital footprint<br/>repos · code · posts · chunks)]
    B -->|RunAnywhere WASM| O[On-device models]
    F -->|static build| R[Render<br/>algsoch.com]

📊 coral_benchmarks — Coral QA Evidence

flowchart LR
    C[catalog.json] -->|required-filters| P[probe_&lt;source&gt;.py]
    P -->|minimal · realistic · edge| S[coral sql]
    S -->|rows + latency| R[reports/ tables + history + findings]
    R --> N[Next.js app]
    N -->|static export| OUT[app/out]
    OUT -->|deploy| LIVE[Live site + GitHub Pages mirror]

Open Source Contributions

Coral MCP — 14 PRs Merged · 20 Open · 6 Approved

Contributor to Coral — SQL-based data abstraction layer for AI agents. Integrated 8 AI providers:

Provider	PR	Description
Voyage AI	#1115	Vector search integration
Sarvam AI	#1112	Indian language TTS/STT
Cohere AI	#1098	Command R integration
Mistral AI	#1011	Mistral model support
OpenRouter	#882	Unified API gateway
LM Studio	#834	Local model serving
Ollama	#798	Local LLM inference
Groq AI	#754	Fast inference provider

Merged: 14 · Open: 20 (6 approved, awaiting merge) · Closed without merge: 2
13 review comments given across 5 reviewed PRs

View all PRs →

Skills & Tech Stack

🤖 AI Engineering

On-Device AI

LLM & Agents

Model Ops

Vision & Speech

🛠️ Development

Languages

Mobile

Frontend

Backend

☁️ Infrastructure

Orchestration

Data

Deployment

✍️ Writing

⚙️ IDE & Tools

AI Coding

Code Editor

Terminal

Other

Languages


Spoken	English, Hindi
Programming	Python, Kotlin, JavaScript, TypeScript, SQL

GitHub Stats

📊 Accounts

Account	Repos	Stars	PRs Merged	Contributions
@FiscalMindset	56	6	170+	—
@algsoch	104+	24+	14	350+
@blindfold-org	2	—	—	—

🔒 Closed-Source Projects

Category	Project	Description
🛡️ Security & Privacy	Blindfold	TDX enclave secrets wrapper — AI agents never see the keys
🎥 Media & Video	Reelforge · ChitraDub	AI-native video editor · dubbing pipeline in 12 languages
🤖 Agentic AI	agentic_chat · vickykumar	Session history + dashboard · personal AI chatbot
🧪 Data & Analytics	footprint · coral_benchmarks	Local-first digital-footprint analytics · Coral query benchmarks
🛠️ Developer Tooling	scrapper	Scrape-Verse Sentinel — self-healing web context for coding agents
🌐 Web & Community	ghevra	Community transparency portal — EN / Hindi / Haryanvi

Featured Work Highlights

Project	Impact	Link
🛡️ Blindfold	AI agents never touch the API keys they use — TDX enclave	Live
🤖 algsoch (AI)	Personal AI chatbot trained on your digital footprint	Live
🎬 ChitraDub	YouTube URL / upload → dubbed MP4 + SRT in 12 languages	GitHub
🎥 Reelforge (video)	AI-native video editor — timeline project model = source of truth	GitHub
📱 algsoch Android	100% offline AI, 7 learning modes, RunAnywhere SDK	GitHub
📺 algsochnews	5-agent pipeline → broadcast video from any article URL	Live Demo
🏥 careops	9 data sources joined via Coral SQL for family care coordination	GitHub
📊 coral_benchmarks	Per-spec query benchmarks for every Coral source	Live

Let's Connect

Platform	Badge
LinkedIn
Discord
Medium
Kaggle
YouTube
Portfolio
Email

MIT License · Built by Vicky Kumar · Render & Vercel