We measure against the humans doing the job today
Every build starts by timing and costing the process as it runs now. If the system cannot beat that baseline on a number you already track, we say so before the budget is spent.
ABOUT THE COMPANY
We moved from building software to operating it.
The Reciprocal Solutions began in Coimbatore as an IT services firm — sites, apps, campaigns, hosting. That work showed us exactly where companies lose their hours. AI is how we give them back.
- 9 yrsBuilding softwareFrom startup to enterprise delivery
- 120+Projects deliveredAcross AI, web, mobile and analytics
- 40+Enterprise clientsRetail, manufacturing, healthcare, BFSI
Where this came from
We followed the work, not the trend.
Every move from IT services to AI agents was forced by a client problem we could not solve with the tools we already had.
The company started with the work businesses around Coimbatore needed doing: a website, an app, a campaign, a hosting bill somebody had to answer for. We took the brief, shipped it, and moved on to the next one.
Then clients started asking us to stay. A retail group wanted its footfall numbers every morning instead of once a quarter. A finance team wanted its invoice queue cleared before anyone arrived at a desk. Those were not design problems. They were operations problems, and software was only part of the answer.
Working on them changed what we measured. A site is finished when it launches. A system that clears a queue is only finished when it is still clearing it nine months later, at a cost somebody signed off on.
So we rebuilt the company around the second definition. Computer vision first, because the cameras were already on the wall. Then retrieval systems over the documents nobody had time to read. Then agents that call the tools themselves rather than drafting a reply for a person to paste.
We kept the rest — web, apps, growth, hosting — because an AI system nobody can log into, find or pay for is not a system. It is a demo.
A demo answers a question. A system absorbs a job.
- Based in
- Coimbatore, Tamil Nadu
- Delivery
- Remote-first, onsite for discovery
- Disciplines in-house
- Four
- Services offered
- 14
What we believe
Five things we will not negotiate.
Several of these cost us work at the pitch stage. They are also the reason the systems we ship are still running a year later.
Reversible actions first
An agent earns its permissions in order: read, then draft, then act. Anything touching money, customers or records stays behind an approval gate until the evaluation history justifies opening it.
The evaluation set is the deliverable
A demo proves a model did something once; an evaluation set proves it does the same thing on Monday. You own that set, and a regression blocks our deploy the way a failing test blocks a merge.
We stay after launch
The interesting failures arrive in month three, when the data drifts and the edge cases surface. We price for that period rather than handing over a repository and wishing you well.
Your data and your IP stay yours
We deploy into your cloud account, use zero-retention model configurations, and assign the code to your company. Nothing you hand us is used to train a public model.
How we are set up
Four disciplines, one roof.
AI programmes stall in the gaps between suppliers — the integration nobody owns, the login that was never built, the page an answer has to appear on. We removed the gaps by keeping the teams.
AI Core
Strategy, agents, retrieval and vision. This team decides what gets automated, what stays with a person, and which number the system is judged on.
- AI Consulting
- AI Agent Development
- AI Automations
- Custom LLM & RAG Systems
- Computer Vision & Analytics
Product Engineering
Platforms, APIs, apps and the interfaces people actually use. An agent with no front door and no audit trail is a script with ambitions.
- Software Development
- Web Development
- App Development
- New Product Development
Growth
Search, social, ads and design. The team that builds the system also makes sure the customers it serves can find you in the first place.
- SEO
- Social Media Marketing
- Paid Ad Campaigns
- Graphics & Brand Design
Infrastructure
Cloud architecture, hosting, model serving and monitoring. Somebody carries the pager for what we shipped, and it is us.
- Cloud, Hosting & MLOps
One backlog
Model work, interface work and infrastructure work sit in the same sprint, so nothing waits on another supplier's release cycle.
No handover
The people who scoped the agent write the code and answer for it in production. Nobody re-explains the business to a second company.
One contract
Discovery, build, hosting and support are priced together, which makes the true cost of running the system visible from week one.
How we got here
Four eras, and what each one taught us.
Not milestones on a wall. These are the four ways this company has earned its money, and the lesson that ended each one.
- SERVICES ERA
Take the brief, ship the brief
Websites, apps, campaigns and hosting for businesses around Coimbatore. We were good at delivery and blind to what happened afterwards — the client's process usually stayed exactly as slow as it had been.
Shipping is not the same as changing the work.
- Delivery discipline
- Every stack a client already owned
- Client-side politics
- PRODUCT ERA
Own the thing, then run the thing
We began building and operating our own platforms rather than only other people's briefs. Uptime, support queues and infrastructure bills became our problem, which changed how we designed everything that followed.
You only learn a system's real cost by paying it every month.
- Operations and on-call
- Cost per transaction
- Roadmaps instead of projects
- AI ERA
Models in production, not in notebooks
Computer vision on cameras clients already owned, then retrieval systems over documents nobody had time to read. The hard part was never the model. It was data access, evaluation, and the week accuracy quietly dropped.
Without an evaluation set, an AI system is only an opinion.
- Edge inference
- Retrieval and grounding
- Regression gates on deploys
- AGENT ERA
Software that finishes the task
Agents that hold context, call the client's tools and complete work end to end, with approval gates on anything sensitive. We build them, score them weekly against the set, and operate them beside the client's own team.
The unit of value moved from the answer to the finished job.
- Tool-use architectures
- Approval and rollback
- Weekly evaluation reporting
Ways of working
Three shapes, and one rule about ownership.
Nearly every engagement opens with a fixed-fee sprint and only becomes open-ended once something is live and earning its keep.
Discovery sprint
We sit with your operators, map the workflow, and score each candidate on value, data readiness and risk. You leave with a ranked backlog and a cost model whether or not you build with us.
- Use-case scorecard
- Data readiness report
- Costed roadmap
Fixed-scope build
One agreed system, one evaluation set, one date. Scope changes are quoted as changes rather than absorbed quietly and delivered late.
- Working system
- Evaluation set you own
- Runbooks and enablement
Retained operation
Monitoring, cost and quality reporting, retraining against fresh feedback, and a named channel for your team. It cancels on a month's notice, because a retainer nobody can leave stops being earned.
- Support SLA
- Monthly eval report
- Roadmap review
Remote by default, onsite where it counts
Discovery happens in your building, with the people who do the work — you cannot map a queue over a video call. Delivery then runs remotely from Coimbatore, with overlap hours agreed at kickoff rather than assumed.
Ownership, in plain terms
- IP transfers to you
- Code, prompts, evaluation sets and infrastructure definitions are assigned to your company on final payment for the phase.
- Your cloud, your keys
- We deploy into your accounts wherever your policy allows, so access can be revoked without our involvement.
- Data stays where it started
- An NDA before anything moves, zero-retention model configurations, and self-hosted open-weight models where the rules require them.
Working here
How we hire, honestly.
We do not keep a standing list of open roles. We do read every application, and we reply to the ones that show the work.
What we look for
- Evidence over vocabulary
- Show us something you built and tell us what broke in it. Which framework you used matters far less than what you learned when it failed.
- Comfort with unclear problems
- Most briefs arrive as a complaint rather than a specification. The job begins by turning one into the other.
- Writing
- Scope, evaluation criteria and handovers are all written here. Clear writing is not a bonus skill on this team.
- Ownership after launch
- You will still be answering for what you shipped in month six. People who like that tend to stay.
The interview, start to finish
- 01
Application
A short note and one piece of work — a repository, a product, a campaign, a design file. No cover letter template required.
- 02
Conversation
Forty-five minutes on what you have built and what you would do differently now. Your questions get answered in the same call.
- 03
Paid exercise
A scoped task close to the real work, timeboxed and paid. You keep the work regardless of the outcome.
- 04
Team session
You meet the people you would sit with and walk them through the exercise. A decision follows within a week.
Tell us which discipline interests you and attach one thing you made. We reply to applications with work attached.
Send us your workinfo@thereciprocalsolutions.com
Next step
Bring us the workflow nobody wants to own.
Forty-five minutes with the people who would build it. You leave with a ranked list and a rough cost, not a proposal deck.