OUR WORK

Conscious products,
built with intention.

Four builds with founders who are in it for the long run. The product, the stack underneath it, and what happened after handover.

TFG × ManoBuddy

A clinical companion for therapists: case management, AI-assisted clinical notes, and case conceptualisation maps.

RoleProduct strategy, design & development
ScopeZero to one, multi-agent AI
StatusIn production, founder-run
SpaceMental healthcare
THE CLIENT Founder-led mental health. ManoBuddy is a founder-led mental health company helping therapists deliver quality care online. They came to us to build the platform that makes that possible.
WHY WE SAID YES Values, first. Better access to mental healthcare sits inside our mission of building for conscious businesses. Sandeep's intention was clear from the first call, and it shaped every decision after it.
WHAT WE BUILT A right hand for the therapist. A browser extension listens quietly in the background and produces clinical notes in proper clinical formats. Conceptualisation maps, per session and per client, give a living view of how a case unfolds. A client-facing platform ties intake, sessions, notes and follow-through into one continuous flow.
UNDER THE HOOD Clinical-grade from day one. A multi-agent AI copilot runs underneath. Session audio and voice reflections become structured notes in SOAP and DAP formats, and OCR reads scanned and handwritten notes. A model-agnostic LLM layer lets us swap the model with no rewrite, and a self-hosted model runs on RunPod serverless GPUs, so sensitive clinical data stays on infrastructure we control. A single-pass transcription fix roughly halved AI spend. PII-safe, encrypted and audit-ready. Turborepo, React and Vite, FastAPI with Celery, Razorpay billing, and CI with Playwright tests.
OUR ROLE Product, end to end. We shaped the direction with the founders, designed the experience, and built the full stack that runs it. The therapist stays fully present and the paperwork writes itself.
THE HANDOVER Built, then handed over. Then we taught the founder to run and extend the product himself. ManoBuddy now runs in production as a one-person company, and we stay on as its technical partner.
“Aish and Krishan make a formidable duo. While the technical expertise of Aishwary is par excellence, Krishan complements it with founder level product thinking and UX sensitivity. Together, they were instrumental in laying a solid foundation for ManoBuddy.” Sandeep · Founder ManoBuddy, IIT Delhi, ex-Founder Peak Mind

TFG × Myntra & Cult.fit Founder

An Indian spirituality platform, taken from a founder's proof of concept to a product built to serve a national audience.

RoleProduct strategy, design & development
ScopePOC to production platform
StatusPre-launch, confidential
SpaceIndic spirituality
THE ENGAGEMENT From a proven POC to a product built for scale. A founder of Myntra and Cult.fit came to us with a strong proof of concept and a working MVP for a spirituality-focused consumer product. He wanted to turn it into a real product that could serve a national audience, so we kept what worked, rebuilt what needed it, and re-architected the platform for scale.
WHY THIS PARTNERSHIP Rooted in intention. Technology rooted in Indian spiritual tradition sits directly within our mission. A founder who has built category-defining consumer products, and who cares about the inner life, made this a natural fit.
WHAT WE BUILT A 24-hour digital ashram. Live and replayable classes in yoga, meditation, bhakti and scripture. Today, Sadhana, Library and Darshan sections. An AI guide grounded in scripture, Hinglish-first with full Hindi and English, subscriptions, mobile OTP login, WhatsApp reminders, 24-hour replays, and a 14-day trial journey for new members.
UNDER THE HOOD Built for scale. A model-agnostic AI layer, the same approach we run at ManoBuddy, with careful memory management so the scripture guide stays coherent and grounded across a long relationship with each member. Next.js, FastAPI, Postgres and Redis, media on S3 and CloudFront, running on AWS behind Cloudflare, with encrypted secrets and a hardened CI pipeline that promotes from dev to staging under branch protection.
SECURITY, HARDENED A breach, caught early. Early on, a staging server was compromised through a known OpenSSH vulnerability. We caught it, remediated, rebuilt on a clean environment, and locked the pipeline down. Security stayed part of the build from that point on.
OUR ROLE From concept to build. Product strategy, product design, and full-stack development, from his early proof of concept through to what is being built today.
PACE A single overnight build to a live preview. Name, brand, nine mockups and a live authenticated preview shipped in one overnight build. The full platform was scoped at three months and delivered in about one. It is now in a pre-launch state, handed to the client's team for a nationwide launch.
CONFIDENTIALITY Held with care. The product is unreleased, so the detail stays in conversation.

TFG × IIT Guwahati

A five-day, in-person AI-assisted coding workshop at IIT Guwahati's Technology Incubation Centre, for government PSU engineers with 2 to 20 years of experience. We skipped the theory and built working skills in the room, on the systems they run every day.

RoleProgramme design & delivery, lead trainer Aish
Scope5-day in-person workshop, 12+ tools
StatusDelivered, June 2026
SpaceGovernment PSU / critical infrastructure
THE MANDATE Capability, on their own code. These engineers write and maintain software that keeps critical national infrastructure running. They wanted their people building with AI across the real software lifecycle, safely, on the code and data they already own. The goal was working capability they could use the next day.
THE APPROACH Foundations first, then AI end to end. We started with the fundamentals: Git, SDLC, and how LLMs actually work under the hood. Then we put AI to work at every stage of the software lifecycle, from requirements through to deployment. More than 80% of the five days was hands-on, on the systems they run every day, across 12+ tools.
THE LINE WE DREW Speed with verification. Engineers who write software for critical infrastructure cannot vibe-code their way through production. We spent time on where speed helps and where verification has to come first.
THE OUTCOME Working software, deployed. By the last day, every engineer demoed a working product connected to real live data, and deployed it. The engineers with twenty years behind them moved fastest: they already knew what good engineering looks like, and AI just removed the typing.
“AI is the junior. You own the output.” The rule we repeated every day at IIT Guwahati

The five days, end to end:

DAY 01 Foundations Git, SDLC, how LLMs work under the hood, and context engineering.
DAY 02 AI-native IDEs & agents AI-native IDE workflows and agentic coding with Claude Code, Cursor and Copilot agent mode, and UI from a prompt.
DAY 03 Full-stack with AI Full-stack product development, building real features end to end.
DAY 04 Data, automation & local LLMs Data analytics, workflow automation, and running private, on-prem models for security and cost.
DAY 05 Production & capstone Production readiness, security, and live capstone demos on real data.
Aish beside the AI-Assisted Coding Training Programme banner at IIT Guwahati
Aish coaching engineers through a hands-on lab at IIT Guwahati
The full training cohort at IIT Guwahati with the programme banner

TFG × Confidential

An AI-guarded vault for premium digital assets: private access in, nothing copyable out.

RoleSecurity architecture & development
ScopeAI protection layer
StatusDelivered, client confidential
SpaceDigital asset security
THE PROBLEM Sharing was the risk. Our client's most valuable work lived in digital assets that had to be shared to be sold, and every share was a chance to lose them. Standard portals only stop honest people.
WHAT WE BUILT A vault that only shows. A private vault where access is granted person by person. Assets render view-only through a delivery layer that never hands the underlying file to the device, so there is nothing to download or copy. Every session carries a dynamic, viewer-specific watermark.
THE AI LAYER AI watching every session. Behind the vault, AI watches how assets are used: access velocity, session patterns, anomalies. Scraping-like behaviour or an unusual session triggers instant lockdown, with an audit-grade trail of who saw what, and when.
THE RESULT Copy-proof by design. The client now sells and shares premium work with named individuals, and it stays theirs.

More projects.

Open Source · Dev Tools FREE
Vibe Testing. A code-aware browser-testing tool that reads your routes and forms, drives them in Playwright, and keeps a regression ledger across releases. Thirteen MCP tools for Claude Code, Cursor and Windsurf. Thousands of downloads, MIT-licensed and free to use. GITHUB ↗
Agentic Pipeline PILOT
Leads Engine. Done-for-you outbound, with you approving every send. An agentic pipeline that researches accounts, drafts outreach, handles replies and books the meeting, with an approval gate at each step. In pilot with a few clients we're co-building with. LIVE ↗
Creator Platform LIVE
Travelling Self. A platform for creators to sell and protect their digital work: subscriptions, access control, and brand identity built in.
TFG Labs 9 MORE
Nine more products. Nine more products we built, for clients, for the open-source community, and for the love of building. They live in TFG Labs. EXPLORE TFG LABS →

Let's explore if we're aligned.

No pitch decks. No sales calls. A real conversation.

TFG Tech for Good · AI product studio Auroville / Tiruvannamalai / Bangalore · working worldwide