Start a project
Product engineering studio · Bhubaneswar, India

Building products.
Solving problems.

We are a ten-person engineering studio in Bhubaneswar. Seven of the projects below are our own products — which is why we can tell you what yours will really cost to build, and to keep running.

We have shipped a voice AI platform, a university ERP, a streaming platform, a CCTV AI layer, a meetings platform and a marketplace.

Start a project See the products
0

Products designed, built and shipped in-house

0

Specialists — engineering, design, QA, DevOps, growth

0

Microservices in our largest client platform

0

Platforms shipped to — web, Android, iOS, desktop

What we do

Eight disciplines, one team.

One team covers the whole path — strategy, design, build, test, deploy, support. No handing your project between three vendors and hoping the pieces fit.

01

Software development

Custom platforms, SaaS, CRM and ERP systems, and the internal tools that quietly run a business.

  • SaaS
  • Multi-tenant
  • CRM / ERP
  • Automation
02

Product development

MVPs for founders and new lines for established companies — scoped to reach real users fast, then scaled.

  • MVP
  • Product strategy
  • Scaling
03

Web development

Portals, marketplaces, e-commerce and progressive web apps that stay fast on Indian networks.

  • Next.js
  • PWA
  • E-commerce
04

Mobile applications

Android and iOS from a single codebase, or native where the product genuinely demands it.

  • React Native
  • Flutter
  • Store releases
05

AI & automation

Voice agents, computer vision, document understanding and LLM features that remove real work.

  • Voice AI
  • Vision
  • LLM agents
06

UI/UX design

Research, wireframes, prototypes and a design system your engineers can actually build against.

  • Figma
  • Design systems
  • Prototypes
07

Cloud & DevOps

Pipelines, environments, monitoring and cost control — so releases are boring, in a good way.

  • AWS
  • Docker
  • CI/CD
  • Cloudflare
08

Consulting & support

Architecture reviews, technical due diligence, and maintenance retainers for products already live.

  • Audits
  • Maintenance
  • Tech partner

Selected work

Seven products. And the part of each one that was hard.

Anyone can list a tech stack. What tells you whether a team can build your product is how they handled the problem with no clean answer. Each project below names its own.

Overview Live product Voice AI

Swaram.

A 24×7 AI voice agent for Indian small businesses.

Swaram answers the phone and the website for clinics, salons, coaching centres and shops — in Hindi, English, Hinglish and regional languages. It answers from the business's own information, books appointments, captures leads, takes orders, and hands difficult conversations to a human with the full context attached. One setup, three surfaces: a phone line, a click-to-talk button on the website, and a face-on-screen video agent. Self-serve onboarding, rupee pricing, UPI billing.

Key features
  • Answers every call, 24×7, in the caller's own language
  • One-click voice button and video agent for your website
  • Books appointments, captures leads and takes orders
  • Warm handover to a human with the whole conversation attached
  • Every conversation saved, searchable, and shown on a live dashboard
The hard part

Real conversations in India switch language mid-sentence. An agent that waits for a full stop, or locks to one language at the start of a call, sounds broken. Natural back-and-forth meant engineering for interruption and code-switching inside a latency budget short enough that the caller never hears the pause — while holding one conversation state across three very different channels.

Real-time voiceLLM agentsTelephonyWebRTCMultilingual ASR/TTSUPI billing
Product gallery
LIVE CALL · 00:42 HINDI BOOKING CAPTURED CONFIRM
Live call, answered in HindiThe agent listens, replies and books the slot while the caller is still on the line.
TALK TO US
A "talk to us" button that talks backVisitors click once and speak. Nothing to install, same agent as the phone line.
Overview In development Multi-tenant SaaS

Edusphere OS.

One platform, from application to alumni.

An AI-native LMS and ERP for Indian higher education. Fifteen functional domains and nine role-based portals cover the full institutional lifecycle — admissions, academics, examinations, finance, hostel, placement, alumni. A Go backend enforces strict tenant isolation on every query, with standardised auth, DPDP-conscious data handling, and feature-flagged modular rollout so an institution adopts at its own pace. On top sits the AI layer: a student companion, moderated discussion rooms and an AI subject agent, designed for integrity and privacy rather than bolted on.

Key features
  • Nine role-based portals across 15 functional domains
  • NEP 2020 and ABC-ready academic structures
  • Accreditation and outcome-based education reporting, automated
  • Dropout early-warning and employability tracking
  • Student AI companion, discussion rooms and subject agent
The hard part

One database serving many institutions means a single missed filter leaks one college's students to another. We made tenancy structural rather than disciplined: every table carries a tenant identifier, every query is scoped through it, and permissions resolve in the API — never in the interface. Modules ship behind per-tenant feature flags, so a college can turn on the AI layer without waiting for the ERP rollout to finish.

GoPostgreSQLMulti-tenantRBACFeature flagsLLM agents
Product gallery
12,480 STUDENTS 94% ATTENDANCE 318 AT RISK PLACEMENT BY DEPARTMENT
Institution dashboardEnrolment, attendance, at-risk students and placement, per department and per campus.
SEM 5 · CS
Student AI companionCourse-aware help with academic-integrity guardrails and moderated discussion rooms.
Overview MVP phase Consumer PWA

Kissa किस्सा.

A home for serialised audio films.

A creator-first platform for episodic stories told as synchronised imagery plus voiceover — horror, sci-fi and thriller first. It sits between the podcast and the short film: cinematic to listen to, cheap enough for one person to produce. Built as an installable PWA on Next.js and Express with PostgreSQL, direct browser-to-storage uploads, and a custom player that keeps images locked to the audio timeline.

Key features
  • Custom player syncing images to creator-set audio timestamps
  • Series, seasons and standalone episodes — grouping is optional
  • Direct-to-storage upload pipeline for large audio and image sets
  • Genre browse, full-text search, watch history and resume
  • Installable PWA with an offline app shell
The hard part

The first data model forced every episode into a series and every series into seasons. It was tidy and wrong — a creator publishing one standalone story had to invent scaffolding to hold it. We rewrote attachment as three modes over nullable relationships, so an episode can stand alone, sit directly under a series, or live inside a season, with navigation bounded to whichever mode it is in. Deciding this before launch avoided a migration across every creator's catalogue later.

Next.jsExpressPostgreSQLAWSCloudflare R2PWA
Product gallery
HORROR S1 · EP 4 · 18:22
The audio-film playerImages advance exactly where the creator set them, with resume-where-you-left-off.
HORROR SCI-FI THRILLER
Browse by genreSeries and standalone episodes side by side, with a persistent mini-player.
Overview Pilot ready Computer vision

Drishti.

Your cameras already see everything. Drishti makes them act.

An AI layer over the CCTV a business already owns, turning passive recording into live alerts. Most cameras write quietly to a recorder nobody watches, so you learn what happened only after the loss. Drishti watches the same feeds continuously and raises a WhatsApp alert the moment something matters. No new cameras, no cloud upload of footage — video stays on-premises, which is both the privacy story and the cost story for Indian factories, campuses and commercial sites.

Key features
  • Intrusion and restricted-zone detection with drawn boundaries
  • Number-plate recognition for gate and yard traffic
  • Fire and smoke detection, well before an alarm triggers
  • PPE compliance and loitering detection on the factory floor
  • Instant WhatsApp alerts with the clip attached
The hard part

An alerting system nobody trusts gets muted in a week, and a site with thirty cameras generates enough movement to produce false alarms all night. The work was in the thresholds — zone geometry, dwell time before loitering counts, and confidence gates per scenario — so that an alert arriving on a supervisor's phone at 2am is worth waking up for. Running inference on-premises against existing analogue and IP cameras added the second constraint: it has to be cheap enough to work without new hardware.

Computer visionEdge inferenceRTSPANPRWhatsApp APIOn-premises
Product gallery
PERSON 0.94 CAM 03 · GATE A OD 02 AB 1234 CAM 07 · ANPR FIRE · ALERT CAM 11 · STORE ZONE · CLEAR CAM 14 · YARD
Live wall, watching for youIntrusion, plates, fire and zone breaches detected across every feed at once.
INTRUSION · GATE A 02:14 AM · CAM 03 VIEW CLIP DISMISS
The alert reaches a humanWhatsApp, with the clip and camera attached — no dashboard to sit and watch.
Overview In development Real-time collaboration

SyncNet.

Unified communication and collaborative architecture.

A single encrypted workspace for high-performance teams — video conferencing, large-scale interactive webinars and persistent team messaging in one place. SyncNet adds advanced screen sharing, real-time participant analytics, and virtual rooms that keep their context between sessions, so a project space still holds its whiteboards and threads when the team returns to it. Built for hybrid and remote work: everything from a two-minute huddle to a global podcast broadcast runs on the same surface.

Key features
  • Instant meetings and HD webinars
  • Podcast hosting and live streaming
  • Encrypted private and group messaging
  • Collaborative virtual rooms that keep context between sessions
  • Advanced screen sharing and presentation tools
The hard part

A huddle for four and a webinar for four thousand are different systems wearing the same interface — one wants everyone publishing, the other wants a stage and an audience. Running both on one architecture meant separating the media path from the room model, so a space can scale from peer conversation to broadcast without moving the participants, the chat or the shared context to a different product.

WebRTCSFU mediaWebSocketsLive streamingE2E messagingAnalytics
Product gallery
● LIVE 4,182 HERE
Premium webinar stageLarge interactive broadcasts with HD video, live Q&A and real-time reactions.
ATTENTION OVER SESSION TALK TIME REACTIONS 73% TOP ROOMS
Engagement analyticsAttention, talk time and reactions across live collaboration sessions.
Overview Live · used daily by our team Web + Windows desktop

NoteX.

An encrypted knowledge base for engineering teams.

A text-first workspace where developers keep notes, code snippets and commands — personally, or shared across nested team workspaces with proper roles. Markdown editor with live preview, autosave, version history, universal search, a command palette, and an optional local AI assistant for titles, summaries and tags. It ships as a web app and as a Windows installer with an embedded server, from the same codebase. We run our own company knowledge on it.

Key features
  • Markdown notes with live preview, autosave and restorable version history
  • Team workspaces that nest without limit, with owner / editor / viewer roles
  • Snippet and command libraries with instant copy
  • Universal search across everything you have access to
  • AES-256-GCM encryption at rest and in transit
The hard part

We encrypt every content field before it reaches PostgreSQL, so anyone holding a database dump sees only ciphertext. That breaks SQL search outright — the database can no longer read its own rows. We moved full-text search into the API process, after decryption, and wrote an idempotent startup migration that encrypts existing rows in place, so live installations upgraded without a maintenance window or a data export.

ReactTypeScriptNode + ExpressPostgreSQLElectronAES-256-GCM
Product gallery
SAVED ✓ # Deploy runbook docker compose up -d npm run migrate Deploy runbook devops aws v14 · 2 min read
Write left, read rightMarkdown with live preview, autosave and every edit snapshotted for restore.
WORKSPACES AES-256-GCM WHAT THE DATABASE SEES WHAT YOU SEE
Nested workspaces, encrypted contentRoles resolve at the root of the tree; a database dump reveals only ciphertext.
Overview Client delivery Microservices

ElektwonikBazar.

A full-scale electronics marketplace, built for a client.

An eight-service marketplace platform covering identity, catalogue, orders, payments, search, notifications and advertising, behind a Next.js frontend of around a hundred pages spanning four distinct experiences — retail buyers, business buyers, sellers and administrators. Prisma and PostgreSQL for data, Redis for caching and sessions, Elasticsearch for catalogue search, Razorpay for payments. Quality is governed by a structured audit plan that walks the codebase in reviewable chunks rather than trusting a single pass.

Key features
  • Eight independently deployable services behind one API surface
  • Four experiences — B2C, B2B, seller console and admin
  • Elasticsearch catalogue search across a large product set
  • Razorpay payments, orders, returns and settlements
  • Seller onboarding, listings, inventory and advertising
The hard part

Eight services and a hundred pages is where consistency quietly dies — the same product shows a different price in three places because each service learned about the change at a different moment. Keeping one truth across service boundaries, and auditing a codebase that size in reviewable chunks instead of one unreadable sweep, was the real work. Not the individual features.

Next.jsNode microservicesPrismaPostgreSQLRedisElasticsearchRazorpay
Product gallery
Components & boards SHOP ALL ₹2,499 ₹899 ₹14,750 ₹5,120
Storefront across four audiencesRetail, business, seller console and admin, sharing one catalogue and one cart engine.
SERVICE TOPOLOGY GATEWAY auth · 4001 catalog · 4002 orders · 4003 payments · 4004 4005 search · 4006 notify · 4007 ads · 4008 POSTGRESQL · REDIS · ELASTICSEARCH · RAZORPAY
Eight services, one contractIndependently deployable services behind a single gateway and a shared data layer.
Overview Under NDA

Client engagements.

Healthcare, education, retail and logistics.

Alongside our own products we build for clients — business platforms, admin and analytics dashboards, e-commerce, mobile apps and AI automation. Most of that work sits behind an NDA, so we do not publish names or screenshots here. We are happy to walk you through the architecture, the trade-offs and the results on a call, and to arrange a reference where the client allows it. The same protection covers you: we sign before we see anything, and IP transfers to you on final payment.

Where we have delivered
  • Healthcare — appointments, records and patient-facing workflows
  • Education — learning portals, admissions and coaching operations
  • Retail — storefronts, catalogue and order management
  • Logistics — tracking, dispatch and operations dashboards
How an engagement runs
DISCOVERY DESIGN BUILD · QA LAUNCH SUPPORT WEEK 0WEEK 2WEEK 4+ GO LIVEONGOING DEMO EVERY 2 WEEKS ONE POINT OF CONTACT NDA SIGNED BEFORE DISCOVERY · IP TRANSFERS TO YOU
The same shape, every projectDiscovery, design, build with fortnightly demos, launch, then support.
Overview Concept · grant stage

Annakshetra & ShramSetu.

Two hardware-plus-software concepts for the MSME Idea Hackathon 6.0.

Annakshetra is a distributed smart-farming network — solar micro cold-storage units and IoT field sensors connecting households directly to small farmer-MSMEs, so demand is decided before cultivation rather than guessed after harvest.

ShramSetu is a workforce operating system for daily-wage workers who do not own a smartphone. An NFC smart card, kiosks and an Odia/Hindi IVR line let a supervisor onboard workers on their behalf. Every job builds a portable skill passport and a verified income record — the path from helper to skilled worker to supervisor to MSME owner, made visible.

Why they are here

These are proposals, not products — we list them because they show where our engineering goes when the constraint is a person without a smartphone or a farmer without cold storage. If your project involves hardware, IoT or field deployment in India, this is the kind of thinking you would get.

IoT sensorsSolar cold-chainNFCKiosksIVR · Odia/Hindi
Concept sketches
COLD 4°C HOUSEHOLD ORDERS DEMAND DECIDED BEFORE CULTIVATION
AnnakshetraSolar cold-storage and field sensors linking households to farmer-MSMEs.
NFC SKILL PASSPORT KIOSK IVR · ODIA/HINDI SUPERVISOR APP INCOME PROOF HELPER → SKILLED → SUPERVISOR → MSME OWNER
ShramSetuSmart card, kiosk and IVR — a workforce system for workers without smartphones.
StartupsHealthcareHigher educationE-commerce LogisticsFinTechManufacturingReal estate Media & creatorsProfessional services

Expertise

What we are unusually good at.

01

Security architecture

Encryption at rest and in transit, hashed credentials, stateless sessions, permissions enforced server-side. We design for the day the database leaks, not just the day of the demo.

02

Multi-tenant systems

One platform serving many organisations without leaking between them. Tenancy is structural — carried on every table and resolved on every query — not a rule people are asked to remember.

03

Real-time voice, video and vision

Latency-bound systems: live calls, WebRTC rooms, streaming and continuous video inference. Different domains, the same discipline — nothing may block the frame.

04

Applied AI, not AI theatre

Agents and assistants placed where they remove real work, with fallbacks, human handover and cost control designed in from the first sprint rather than discovered on the first invoice.

05

Products built for India

Indian languages, UPI and rupee pricing, IVR for the phone-less, and interfaces that hold up on a mid-range Android and a patchy connection. Our home market, not a localisation afterthought.

06

Service decomposition at scale

Monolith to services and back again where it makes sense — eight-service platforms with a hundred-page frontend, held together by one contract and one source of truth.

07

Migrations that don't stop the business

Schema changes, data backfills and encryption migrations written to be idempotent and safe to re-run, so upgrades happen without downtime or a weekend anyone dreads.

How we work

A process you can see through.

Seven stages, a working demo every fortnight, and one person you can always call. You will never have to ask us where the project stands.

STAGE 01

Discovery

We map what you need, who it is for and what success looks like — before anyone writes code.

STAGE 02

Product strategy

Scope, milestones, architecture and a written estimate. What is in version one, and what waits.

STAGE 03

UI/UX design

Wireframes to clickable prototype. You approve the experience before it becomes expensive to change.

STAGE 04

Development

Two-week cycles, each ending in a working demo on a staging URL you can open any time.

STAGE 05

Quality assurance

Functional, device and performance testing by a dedicated QA engineer — not by the developer who wrote it.

STAGE 06

Deployment

Cloud setup, CI/CD, monitoring, backups and a documented handover of every credential.

STAGE 07

Support & iteration

A warranty period after launch, then an optional retainer for fixes, improvements and new releases.

Working together

Three ways to engage us.

Fixed scope

Defined project, fixed price

Best when the requirement is clear. Agreed scope, milestones and cost, paid against deliverables. No surprises on the invoice.

Dedicated team

Your engineers, monthly

A pod of designers, engineers and QA working only on your product, month to month. Best when the roadmap is still moving.

Product partner

From idea to launch, with you

We act as your technology team — strategy, build, launch and iteration — for founders without an in-house engineering function.

The team

The people who will actually build it.

No subcontracting, no swap between the pitch and the project. These are the people you will meet on the calls.

APAmitav Pusty

Amitav Pusty

Full-Stack Engineer

React · Node · PostgreSQL
MKMilan Kumar

Milan Kumar

Full-Stack Engineer

Next.js · APIs · Integrations
ASAnish Sharma

Anish Sharma

Senior Backend Engineer

Architecture · Data · Scale
AJAlquama Jia

Alquama Jia

Frontend Engineer

React · TypeScript · UI systems
JSJnanaranjan Sahoo

Jnanaranjan Sahoo

AI & DevOps Engineer

LLM systems · AWS · CI/CD
SNSubrat Nandi

Subrat Nandi

Mobile Engineer

React Native · Flutter · Releases
SDSiddhartha P. Das

Siddhartha P. Das

UI & UX Designer

Figma · Design systems
ASArya Sunirmal

Arya Sunirmal

Tester

Test plans · Automation · Devices
JDJagat Jyoti Dash

Jagat Jyoti Dash

Growth & Marketing Lead

Positioning · GTM · Content

Your pod

Every engagement gets a named delivery lead as your single point of contact, a solution architect who signs off the technical approach, and a QA engineer independent of the developers. You get their direct numbers on day one.

Why Texotic

Reasons that hold up after signing.

We build our own products

Seven of them. We carry the consequences of our own architecture decisions, which changes how we make yours.

You own everything

Code, repositories, designs, credentials and IP transfer to you. NDA signed before discovery. No lock-in, no hostage infrastructure.

One team, whole scope

Design, engineering, QA and DevOps sit together. Nothing is lost in a handover between vendors, because there isn't one.

A registered, structured firm

GST and Udyam (MSME) registered, DPIIT application in progress. Proper contracts, proper invoices, proper documentation.

Visible progress, fortnightly

A working demo and a written update every cycle, on a staging environment you can open yourself. Progress you can check, not just believe.

We stay after launch

A warranty window on every project, then a support retainer if you want one. Launch day is a milestone, not an exit.

Stack

Tools chosen for the job, not the trend.

Frontend

  • React
  • Next.js
  • TypeScript
  • Tailwind
  • Vite

Backend

  • Node.js
  • Express
  • Go
  • TypeScript
  • REST

Mobile & desktop

  • React Native
  • Flutter
  • Electron
  • PWA

Data

  • PostgreSQL
  • MongoDB
  • MySQL
  • Redis
  • Elasticsearch

Cloud & infrastructure

  • AWS
  • Docker
  • CI/CD
  • Cloudflare

AI & real-time

  • LLM agents
  • Voice AI
  • Computer vision
  • WebRTC

Design

  • Figma
  • Adobe CS
  • Design tokens

Ways of working

  • ClickUp
  • Git
  • Staging envs
  • Fortnightly demos

Before you ask

The questions clients ask first.

Who owns the code and the intellectual property?

You do. Source code, repositories, design files, documentation and IP transfer to you on final payment, and we sign an NDA before discovery starts. We keep nothing that would make it hard for you to move on.

How do we start, and how long does a proposal take?

A 30-minute discovery call, then a written proposal within three to five working days covering scope, milestones, timeline, cost and the assumptions behind them. If we need a deeper technical review first, we will say so on the call.

How do you price work?

Three ways: fixed price against a defined scope, a monthly dedicated team, or a longer product partnership. Fixed scope suits clear requirements; a monthly team suits a roadmap that is still changing. We will tell you which one fits your project, even when it is the smaller number.

How will we communicate during the project?

One named delivery lead as your point of contact, a shared board where every task is visible, asynchronous updates through the week, and a working demo at the end of each cycle. Escalations go straight to the leadership team.

What happens after launch?

Every project includes a warranty period covering defects at no extra cost. After that you can take a support retainer for monitoring, fixes and new features, or take the handover and run it in-house — we document it either way.

Can you work with our existing team or an existing codebase?

Yes. We join existing repositories, follow your conventions and work alongside in-house engineers. We also take on rescue work — auditing an inherited codebase, stabilising it, then improving it in a planned sequence rather than a rewrite you did not ask for.

Do you work on AI features, or just conventional software?

Both, and we are careful about the difference. Four of our own products carry AI at their core — voice agents, computer vision, LLM assistants — so we can tell you honestly when a model earns its place in your product and when a simpler system would serve you better and cost less to run.

Where are you based, and does the time zone matter?

Bhubaneswar, Odisha, India (IST, UTC+5:30). We overlap comfortably with Europe, the Middle East and APAC, and keep a fixed morning window for clients in the Americas.

Start here

Tell us what you are trying to build.

Send a paragraph about the idea, or a full requirement document — either works. You will get a reply from an engineer, not a sales script.

Email us your brief Visit texotic.in