Sai Thanush Reddy
Full-Stack Engineer · AI products, real-time systems, cloud infrastructure
“I turn ‘we should build this’ into ‘here’s the production URL.’”
About
I build the hard middle between a sharp product idea and software people can actually rely on. That means product UI, APIs, data models, AI workflows, real-time systems, deployment, and the unglamorous operational details that decide whether a launch holds up.
My recent work sits where products get technically interesting: LLM applications that can prove their answers, voice AI running on GPUs, streaming build systems, and infrastructure that is observable when something inevitably breaks at 2 a.m.
I like simple architectures, boring technology where it fits, and understanding systems deeply enough to build the primitives when it doesn't. The goal is not to use every tool. The goal is to make the product faster to learn from, harder to break, and easier to evolve.
What I Bring to a Founding Team
End-to-end ownership
I can carry a feature from an ambiguous customer problem through product decisions, implementation, deployment, and the first ugly edge cases — without handing it across five specialties.
AI with a reliability instinct
I build AI products with the surrounding system in mind: structured inputs, retrieval that can be inspected, streaming UX, safeguards, and a clear boundary between a useful answer and a confident guess.
Speed that survives contact with users
I optimize for learning velocity, not demo velocity. Ship the smallest credible version, instrument the risky parts, then make the next decision with evidence instead of optimism.
Systems thinking without the theater
Containers, queues, caches, observability, and secure data boundaries are useful when they remove risk. I know when to introduce them — and when a straightforward service is the better product decision.
Selected Work
Resonance
liveA voice AI product built for the entire workflow, not just a model demo: text-to-speech and zero-shot cloning on serverless GPUs, organization-scoped voice libraries, generation history, and secure audio storage. The interesting work was making expensive inference feel like a dependable product.
Next.js · TypeScript · tRPC · Prisma · PostgreSQL · Modal · Clerk · Cloudflare R2
Metadata-Guided RAG
liveDocument Q&A designed around a question founders and users actually ask: “Can I trust this?” It creates a structured metadata layer at ingestion, plans retrieval with SQL and targeted vector search, and verifies every displayed quote against the retrieved source. Deterministic settings, pinned models, and no invented fields.
FastAPI · SQLite · ChromaDB · sentence-transformers · GPT-4o
Beaam
codeAn AI app builder that closes the gap between generated code and a product someone can inspect. LLM output streams over WebSockets into a live, hot-reloading preview, with full-stack generation and one-click deployment built into the loop.
Next.js · TypeScript · FastAPI · PostgreSQL · Redis
Kubernetes Microservices Platform
docsAn 11-service Kubernetes environment that treats “it works on my machine” as the beginning of the problem. Services deploy with Helm and are paired with the operational context needed to debug them: metrics, logs, and distributed traces.
Kubernetes · Helm · Grafana · Prometheus · Loki · Tempo
One-Click Deployment Platform
codeA Vercel-style deployment platform that makes the invisible part of shipping visible: an automated Docker-to-ECR pipeline with build logs streamed to the browser in real time through Redis Pub/Sub and WebSockets.
Node.js · Docker · AWS ECR · Redis · WebSockets
SnapCache
codeA Redis clone in Go: an in-memory key-value store that speaks RESP over TCP and persists data. Built from first principles because the fastest way to gain good systems judgment is to understand what your abstractions are hiding.
Go · TCP · RESP
Technical Range
- Product layer
- React, Next.js, TypeScript, Tailwind CSS — polished interfaces, responsive state, and product surfaces that make technical systems usable
- Application layer
- Node.js, FastAPI, GraphQL, REST, tRPC, WebSockets — APIs and real-time workflows that stay understandable as they grow
- Data layer
- PostgreSQL, MongoDB, Redis, SQLite, Prisma — data models, caching, queues, and retrieval systems
- AI layer
- OpenAI & Claude APIs, RAG pipelines, streaming, embeddings, structured extraction, eval-minded product design
- Systems layer
- Docker, Kubernetes, AWS, Cloudflare R2, Helm, Grafana, Prometheus, Loki, Tempo
- Languages
- TypeScript / JavaScript, Python, Go
Security Disclosures
Orchids
responsible disclosureIndependently identified a critical production-infrastructure vulnerability, reported it responsibly, and coordinated disclosure through to a verified fix. I bring that same adversarial thinking to the products I build.
The Working Style
I’m at my best with teams that want to move quickly without pretending consequences do not exist. Give me a customer pain point, a rough direction, and room to own the details; I’ll turn ambiguity into a deliberate build plan and a working product.
If you are a founder, the useful question is not “Can this person build a landing page?” It is “Can I trust them with the part of the company that does not have a playbook yet?” That is the kind of work I am looking for.