From ETL scripts to founding engineer on a live SaaS product.
I'm Kamlesh — a backend-first engineer based in Bengaluru who's just as comfortable owning an entire product end-to-end. I specialize in Python-based backend engineering, distributed data pipelines, and production AI infrastructure — and as founding engineer on a healthcare SaaS platform I built as a personal project, I took it from an empty repository to real paying customers, alone, before hiring the team that runs it today.
years building production systems
tokens/month in LLM infra owned
a product taken from empty repo to paying customers
My path here
I started at DUIT Technologies writing Python ETL scripts and evaluating machine learning APIs — unglamorous work that taught me the thing that's stuck with me since: most of engineering is making messy, real-world data behave. From there I spent over three years at Hevo Data as a full-stack engineer, where I learned that the line between "backend" and "frontend" matters far less than the line between "shipped and measured" and "not." I owned APIs, migrated a Django site to Next.js, ran A/B tests, and watched Core Web Vitals move because of decisions I made.
In my current role, I architect the LLM infrastructure that processes over 5 billion tokens a month — the kind of scale where a naive design doesn't just run slowly, it becomes unaffordable. Moving inference onto a self-hosted GPU platform and cutting our AI processing bill 62% is one of the projects I'm proudest of, not because of the percentage, but because it's exactly the kind of problem I want more of: real constraints, real budget, real consequences for getting the architecture wrong.
Alongside that, I'm the founding engineer on a healthcare claims platform I built as a personal project — one that didn't exist when I started. No codebase, no architecture, no team. I built the whole thing: frontend, backend, infrastructure, billing, the data pipeline, all of it — then hired and onboarded the engineers who've joined since. It's the clearest proof I have that I can take a product from nothing to production and keep it running once real customers depend on it.
Building a product from zero
If you want the concrete version of "can this person own a whole product" — this is it. Dual-database architecture, a full Stripe billing lifecycle, RBAC, an automated data pipeline with QA gating, and a team I hired and still mentor. Built from a blank repository to a platform real healthcare organizations use daily.
Read the full case studyWhat I actually care about
Deterministic over clever
At the scale I work at — billions of tokens, billions of events — a clever system that's hard to reason about will eventually fail in a way nobody can debug at 2am. I'd rather build something boring that's always right.
Cost is a design constraint
Every infrastructure decision I've made that mattered — Trino over BigQuery UDFs, self-hosted vLLM over hosted APIs — came from treating monthly cost as a real architectural constraint, not an afterthought for finance to worry about.
Own it end-to-end
I'd rather ship a feature from API to UI to the metric that proves it worked than hand off at the API boundary. It's why my personal project exists as a single product I built top to bottom, and why I built every interaction on this site myself.
Comfortable being the first hire
Being founding engineer means there's no one to ask when the architecture doesn't exist yet. I don't just tolerate that — building this personal project from an empty repo to production customers is the work I'm most proud of.
Open to the right problem
I'm open to SDE-2 / SDE-3 backend, platform, or full-stack roles in product-led, AI-driven teams — and to founding/lead engineering engagements for startups and founders who need someone who can take a product from idea to production alone.