Services01Work02Tools03Process04Experience05FAQ06Let's talk07

I buildthe softwareyou're missing.

Vyaas Srinivasan

Product BuilderAI Systems

You work directly with me from the first decision through to something live and holding up.

Book a call See systems I've shipped

First call is free and takes thirty minutes. No deck, no pitch. Prefer WhatsApp? Message me directly.

Build a new product
From a rough idea to a production system real customers can use.
Replace a manual workflow
Turn spreadsheet and WhatsApp coordination into reliable software.
Scale an internal system
Modernize a system that is straining as the business grows.
Vyaas Srinivasan outdoors, smiling at the camera
Billions

Of daily requests and events through infrastructure I've built and run

5

Startups & scale-ups, from eBay to founding engineer

2

Production systems built solo: one enterprise, one consumer

Where I've built
  • eBay
  • Branch Metrics
  • Descript
  • Yuna
  • Weavzi

Where the work actually starts

Deciding what to build
is the job.

01 What I do

What gets better
when we work together.

Faster delivery, fewer manual steps, clearer information, and software your team can rely on every day.

02 Selected work

Two products.
One mine, one delivered.

Weavzi is my own, built to run a factory. Vyoma Veda Astro was built for a founder and takes payments from real customers on the public internet. Both are complete systems rather than features.

Textile ERP · Factory operating system

Weavzi

Solo build: architecture, backend, interface

A textile ERP and factory operating system for garment manufacturing and job-work factories. Not a dashboard bolted onto spreadsheets. The whole operation, from a sales order through supplier movement and the shop floor to the ledger, running as one system. The wider proof is the build itself: translating a complicated operation into software that holds together end to end.

  • Sales
  • Supplier movement
  • Shop floor
  • Inventory
  • Quality
  • Finance
6
Modules in one platform
Demo
Full product, live on demo data
India
Built close to the industry it serves
View build notes
What makes it hard
  • Job-work ledger + supplier movement as a single consistent trail, not disconnected sheets.
  • Offline-tolerant shop-floor entry, reconciled on reconnect.
  • Strict Postgres types, no ORM magic, readable after handover.
Explore the Weavzi textile ERP
Weavzi Command Center showing factory priorities, blocked production operations, and material shortages
Real product interface · Demo data
Customer-facing application

Vyoma Veda Astro

Built for a founder: chart engine, backend, interface

A full consumer SaaS. A classical Jyotish engine computes the real chart; AI writes it up in plain language and cites the exact placement behind every answer. Accounts, memberships, payments and PDF export underneath.

  • Jyotish engine
  • AI chat with citations
  • Compatibility & synastry
  • Accounts & memberships
  • Payments
  • PDF reading export
4
Ways in, from free to full reading
$15+
Paid product, live today
Encrypted
Birth data private by default
View build notes
What makes it hard
  • Deterministic Jyotish engine in code, then AI only for phrasing. Every sentence cites the exact placement.
  • E2E encrypted birth data, membership + payments + PDF export that actually works in production.
  • Eval harness over real charts before any copy change ships.
Visit vyomavedaastro.com
Vyoma Veda Astro answering a question about the current dasha in plain language, with a 'Why this appears' citation and chart-specific follow-up questions
Real product interface · Live today

03 How we'd work

Low friction to start.
Visible progress throughout.

You should know within one conversation whether this is worth your time, and see working software long before the invoice feels large.

You can start as small as you like
  • A second opinionDays, not months. A review of your architecture, a build-vs-buy call, or a direction checked before you commit budget to it.
  • A first milestoneOne useful thing, live. Fixed scope and a written number agreed before work starts, so the commitment is bounded.
  • The whole systemIdea to live product, with me owning delivery end to end. The Weavzi and Vyoma builds below are both this shape.
Step 01

A short conversation

Thirty minutes on what you're trying to change and what's in the way. No deck, no pitch. You leave with a straight answer on whether I can help.

Free · 30 min
Step 02

A scoped plan

A written proposal: what gets built, in what order, what it costs and what it should change for the business. Fixed scope for the first milestone.

Within a week
Step 03

Shipped in slices

Working software in your hands early and often, with a weekly rhythm of demos and decisions. No six-month black box.

Ongoing
Step 04

Kept running

Launch is the middle of the job, not the end. I stay on to fix what only shows up in real use, and to build the next thing the system turns out to need once people depend on it.

After launch
The part nobody likes asking about

Nothing about the money
arrives as a surprise.

I don't publish a rate card, because what a piece of work costs depends on what it is. What I can tell you up front is that you will never see a number you haven't already agreed to.

  1. 01A written price before anything startsScope, milestones and a number, agreed in writing. No work begins, and nothing is billable, until you've said yes to it.
  2. 02The first milestone is fixed priceYou commit to one bounded piece of work, not to an open-ended engagement that grows an invoice each month.
  3. 03You can stop after itIf the first milestone doesn't convince you, that's a clean place to end. You keep the code, the repository and the documentation regardless.

One person to deal with.
As many as the work needs.

On most projects I design and build it myself, which is why it moves quickly. When the work is bigger than one pair of hands, I lead the team that does it — that is the job I've held as a lead and founding engineer, on systems handling billions of events a day. Either way you deal with me, and I own delivery.

  • Your code, your repositoryEvery commit lands in a repo you own from day one, not handed over at the end of the engagement.
  • Contract, NDA and IP assignment as standardA written agreement before work starts, an NDA whenever you want one, and the intellectual property assigned to you. Not extras you have to negotiate for.
  • Documented as it's builtArchitecture notes, setup instructions and the reasoning behind each decision, written for whoever picks it up next.
  • No proprietary lock-inStandard frameworks and standard hosting. Nothing that only I can run, and nothing you have to keep paying me for.
  • Honest availabilityYou'll know my capacity and other commitments before we start, not after you're depending on me.
  • Interface includedYou don't need to hire a designer before anything can start. The interface is built as part of the system, clear and consistent and meant to be used rather than admired. If what you need is brand and visual identity work, that's a specialist job and I'll tell you so.
  • Built for sensitive dataAccess control, audit trails and encryption designed in from the start. I've built to HIPAA and SOC 2-aligned standards for health data before.
  • Set up for teams in other timezonesNine years at eBay, Branch Metrics and Descript in the US means the review culture, the planning rhythm and the standard a piece of work is held to are already familiar. I work from a different timezone to most of my clients now, and set the overlap deliberately at the start rather than leaving it to be discovered in month two.

04 Experience

Built in startups.
Tested at scale.

Across AI products, large-scale backend systems, consumer software and internal enterprise tools, built for companies on both sides of the world, which is mostly useful because it means fewer surprises.

01
WeavziCoimbatore, India
Factory OS, complete build

Lead Engineer / Product Builder

02
YunaSan Francisco · remote from India
58% latency cut · 99.2% uptime

Founding Engineer, AI & Backend

03
DescriptSeattle, USA
Production AI on GPU clusters

Senior Engineer, AI Infrastructure

04
BranchSeattle, USA
20B daily events · $500K+ saved

Senior Engineer, Systems & Scale

05
eBayBellevue, USA
1.5B daily requests, sub-ms

Software Engineer

What the AI work
actually involves.

Most of it isn't the model. It's the retrieval, the guardrails, the evaluation and the infrastructure that keeps the thing standing on a Tuesday afternoon.

  • Retrieval that cites its sourceRAG pipelines over pgvector where latency is treated as a product requirement, not an afterthought. Twelve seconds down to five on a live assistant.
  • Agents with guardrailsAgent-based safety rails for high-stakes conversations, on an architecture aligned to HIPAA and SOC 2, holding 99.2% uptime through daily releases.
  • Evaluation before deploymentA custom evaluation harness for A/B testing prompts and models, so a change ships because the numbers moved, not because the demo felt better.
  • Inference that stays upProduction AI on GPU clusters with Temporal orchestration, OpenTelemetry tracing and Kubernetes underneath. The unglamorous half of shipping AI.
Education
  • M.S. Computer ScienceNortheastern University · Boston, USA
  • M.S. Software EngineeringCoimbatore Institute of Technology · India

Before you take my word for any of it

Nothing on this page
needs taking on trust.

Every figure quoted here comes from work I can walk you through in detail on a call. If a number on this page matters to your decision, ask me about it and I'll show you where it came from.

05 Before you write

The questions
people hesitate to ask.

If something here is still unclear, that's a good enough reason to get in touch on its own.

What does a project cost?

It depends on what you're building, but you won't be surprised. You get a written number for a defined scope before any work starts. If your budget and the problem don't match, I'll tell you early rather than after you've spent.

When could we start?

It depends on my current capacity and the size of the first milestone. After the initial conversation, I'll tell you honestly when I can begin and include the start date in the proposal. I don't accept work and quietly leave it waiting in a queue.

What if my project is small?

Small is usually how these start. The first milestone is deliberately contained — one process, one system, one decision — because it's the cheapest way for both of us to find out how the other works. Most engagements grow from there, once the first piece is running and the next bottleneck is obvious. Some of the most useful things I've built began as one annoying process someone finally wanted to stop doing by hand. What doesn't fit is a one-off afternoon's fix: not a bad thing to need, just not what I'm for, and I'll point you somewhere better.

Who owns the code and the intellectual property?

You do, from the first commit. Work happens in a repository you own, the IP is assigned to you in the contract rather than licensed to you, and I'll sign an NDA whenever you'd like one.

What if it isn't working out?

The first milestone is fixed in scope and price precisely so that ending there is a clean, cheap decision rather than an argument. I'd much rather you stop after one milestone than spend six months regretting the third.

How quickly will I see something working?

Most first milestones land in weeks rather than months, and that's deliberate. You should see working software early enough to change your mind while changing it is still cheap.

How do you use AI while building software?

I use AI to accelerate research, implementation, testing, debugging and documentation. Architecture, security decisions, code review and accountability remain mine. It helps me deliver more quickly; it doesn't replace engineering judgment or become an excuse for code nobody understands.

Can you take an idea to a launched product?

It's a large part of the work. Weavzi and Vyoma Veda Astro were both taken from nothing to complete, working systems, including the parts founders tend to discover late: accounts, permissions, payments and the operational tail behind the demo.

What if what we have is breaking under load?

That's a different job from building new, and it's one I've done for years. I've run production systems handling 20 billion events a day and 1.5 billion requests a day, and cut response time on a live AI product from twelve seconds to five. The first step is almost always measurement, not a rewrite — most systems are slow somewhere nobody has looked.

Can you lead a team, not just write the code?

Yes. I've worked as a lead engineer and a founding engineer, which means architecture and delivery and the people side of it, not just the commits. On smaller projects I build it myself because that's faster. On larger ones I bring in and direct engineers I've worked with before, and you still have one person to talk to.

Why not hire a full-time engineer or agency?

If the work is permanent, continuous and central to your company, building an internal team may be the right decision. I make sense when you need senior execution now, before a full team is justified, or when one important system needs focused ownership without a long hiring process. You work directly with the person who designs it and answers for it, not with an account manager relaying to someone you never meet.

Do you work with manufacturing businesses?

Yes. It's some of the work I know best. Weavzi is a full factory system I built for textile and job-work manufacturing, so terms like job-work billing, subcontractor movement and shop-floor tracking don't need explaining from scratch.

What if I already have a developer or a team?

Then what you probably need isn't more hands, and that isn't what I'd be. It's usually one of two things: a decision nobody has the context to make yet, or a system nobody on the team has the room to own. I'll work alongside them on either, and say so plainly if the honest answer is that you just need to hire.

Can you handle sensitive or regulated data?

Yes. I've built production systems on HIPAA and SOC 2-aligned architecture for health data, where access control, encryption and audit trails aren't features you add later. If your data carries obligations, tell me early and they shape the design rather than getting bolted on at the end.

Do you work with clients in other countries?

Yes. I spent nine years in the US building for eBay, Branch Metrics and Descript, so I know from the inside how teams there expect to work: the review culture, the planning rhythm, the standards a piece of work is held to. I deliver to teams several timezones away, which is how Yuna was built. Tell me your working hours early and we'll set the overlap deliberately rather than discovering it in month two.