Skip to content
Sriram Nyshadham
Fig. 01 · PortfolioRev. 2026-10

Sriram
Nyshadham

Portrait of Sriram Nyshadham

full stack AI engineer · systems

I built a live ambient scribe and 30+ AI agents that clinicians use every day.

Open to Full Stack and AI engineering roles in AI-enabled SaaS, healthcare and developer tools.

  1. 01audio
  2. 02ws stream
  3. 0330+ agents
  4. 04clinician review
  5. 05record
Fig. 1 How a visit becomes a clinical note: audio streams over WebSockets to 30+ agents, and a clinician reviews every draft before it reaches the record.

Ask the assistant

Hi 👋 I'm Sriram's AI assistant. Don't just read the resume — ask me anything about his experience, projects, or how he works. I can even pass a message along to him.

Fig. 02 · Impact
AI agents in production, with 200+ tools
30+
infrastructure cost cut, Kubernetes to serverless containers
~50%
successful production deploys since 2024
10,000+
pull requests reviewed as primary reviewer
840+
Fig. 03 · Case study

A live ambient scribe for clinical visits

Writing clinical notes pulls a clinician's attention away from the patient. The scribe drafts the note during the visit, so the clinician reviews a draft instead of writing from scratch.

It runs on the platform I lead at Atelia: visit audio streams over WebSockets into a set of specialized agents, and nothing reaches the clinical record until a clinician has reviewed it.

Read the case study
  1. 01ProblemWhy notes are the bottleneck
  2. 02ArchitectureFrom visit audio to the record
  3. 03TradeoffsCheckpointed state, clinician review
  4. 04ResultsWhat changed in production
Fig. 04 · Projects

Things I've designed and shipped on my own

  • Evenly

    Live

    AI grocery bill splitter

    Exact to the cent, every time.

    Expo / React Native / Convex / Better Auth / Vertex AI · Gemini / TypeScript

    How it works

    Scan a grocery receipt, split it item-by-item, and share each person's exact total. Gemini 2.5 Flash on Vertex AI extracts the line items into a structured schema; a pure, unit-tested, integer-cent split (largest-remainder) guarantees every per-person total sums back to the receipt — no lost or extra pennies. Native iOS + Android via Expo on Convex, with Better Auth and a biometric launch gate.

  • peekdiff

    Live

    Private-repo diff and PR review

    Your private code never touches my server.

    Next.js 15 / React 19 / GitHub OAuth / REST + GraphQL / TypeScript

    How it works

    A fast, virtualized GitHub diff and code-review tool for private repositories, built on the open-source DiffsHub viewer. I built the auth and review layers on top: private-repo access via a GitHub OAuth App where diffs are fetched client-side straight from the GitHub API, so private source never touches the server; a full inline review flow (threaded comments, GitHub-flavored markdown, batched Approve / Request-changes / Comment submit); and per-file and per-folder 'viewed' tracking synced both ways with GitHub over GraphQL. Built on Next.js 15 + React 19.

  • Cipher

    Live

    End-to-end encrypted chat

    Even the server can't read your messages.

    Next.js / Convex / Better Auth / Web Crypto / RSA + AES-GCM

    How it works

    Real-time 1:1 chat with true end-to-end encryption — a fresh AES-GCM key per message, wrapped to each participant's RSA key, all generated in the browser via the Web Crypto API. The server only ever stores ciphertext (every message has a 'show ciphertext' toggle to prove it). Built on Next.js + Convex for realtime, with Better Auth (Google/GitHub).

  • Quill

    On-device writing assistant for macOS

    Fully on-device, no network.

    Swift / SwiftUI / Apple Foundation Models / Accessibility API

    How it works

    A Grammarly-style writing assistant that runs entirely on-device via Apple's Foundation Models framework — no accounts, no API keys, no network calls. An AXObserver tracks the frontmost app's selection and typing through the macOS Accessibility API; the model returns exactly one corrected string via guided generation and is never trusted with character offsets, so a local word-level LCS diff computes the real edit and highlights it in a non-activating floating panel that never steals focus. Works across native apps, Electron apps, and browsers, with a clipboard-paste fallback for apps that reject direct Accessibility writes.

  • Penny

    Notes-first expense tracker

    Log expenses like writing a note.

    Expo / React Native / SQLite · Drizzle / AWS Lambda / Vercel AI SDK / Claude on Bedrock

    How it works

    A local-first expense tracker where you log spending the way you'd jot a note: type "coffee 6.50, uber to airport 32" and Penny turns it into structured, categorized expenses. A parser service uses the Vercel AI SDK's schema-validated structured output (generateObject + zod) with a swappable model (Claude Haiku 4.5 on Bedrock by default), shipped as a single AWS Lambda. The Expo app stores everything on-device with SQLite + Drizzle and falls back to an offline parser, so it works with zero setup.

  • This site

    Live

    AI portfolio

    The assistant above answers from my resume with RAG.

    Next.js / RAG / Vertex AI / Pinecone / Vercel AI SDK

    How it works

    The AI assistant on this page answers questions about my work using RAG: Pinecone vector search over my resume with Google Vertex AI (Gemini) for generation, plus an agentic tool that can email me directly from the chat. Built on Next.js 15 with streaming responses and rate limiting.

  • Cotton leaf classifier

    Deep-learning disease detection

    IEEE-published and cited by 27 papers.

    Python / Deep Learning / Computer Vision / Transfer Learning

    How it works

    Deep-learning computer vision for cotton-leaf disease identification. Benchmarked four transfer-learning CNNs (ResNet50, VGG19, InceptionV3, ResNet152V2); ResNet152V2 reached 98.36% accuracy. Published at IEEE (ICISSGT 2021) and cited by 27 papers.

Fig. 05 · Experience

Experience

  1. Aug 2023 – Present

    Lead Full Stack Engineer

    Atelia Software Inc. · Healthcare technology platform

    Lead engineer on a HIPAA-compliant clinical platform used daily by doctors, hospitals and patients, from real-time clinical AI to ~50% cloud cost savings.

    Show details
    • Real-time Clinical AI. Streamed live doctor-patient sessions over native WebSockets with AI-driven clinical suggestions; shipped LLM chat, a live ambient scribe, and RAG-based documentation agents used daily in production.
    • AI Pipelines. Built 30+ specialized agents and 200+ tools with multi-agent routing, checkpointed state and clinician review, on top of prompt engineering, embeddings, and pgvector retrieval.
    • Full-Stack & Mobile. Own the backend (24 services, ~950 APIs) plus 9 web apps and 3 mobile apps, built with React, React Native, Node.js, and NestJS; led the Flutter → React Native migration (2025) and shipped apps to the App Store and Play Store.
    • Cloud & Cost Optimization. Moved containerized microservices from Kubernetes to serverless containers, cutting infrastructure costs ~50%; infrastructure as code with Terraform/OpenTofu, managed PostgreSQL, IAM, and automated GitHub Actions CI/CD.
    • Reliability & Mentoring. Built fault-tolerant services (structured logging, retries, graceful degradation), tuned PostgreSQL for low-latency RAG, and mentored engineers on code review and CI/CD practices; top contributor among 38 engineers and primary code reviewer for 840+ pull requests.
  2. Jun 2021 – Jul 2022

    Full Stack Developer (Contract)

    Atelia Software (India) · via KK Consultancy

    Built the first Atelia Health mobile apps for remote patient monitoring.

    Show details
    • Built the first Atelia Health mobile apps in Flutter for iOS and Android: remote patient monitoring, video calls and secure messaging.
    • Integrated Bluetooth Low Energy vitals devices (blood pressure, heart rate, temperature) so readings sync to clinicians in near real time.
    • Designed Firestore data models and security rules with role-based access to HIPAA-sensitive records, plus Cloud Functions for threshold alerts.
    • Shipped the Doctor and Patient apps to the App Store and Google Play and supported the first clinical users.
  3. Mar 2020 – May 2021

    Full Stack Developer (Contract)

    Eficens Systems LLC · via KK Consultancy

    Built FLAIR, a full-scale staffing CRM, from the ground up: payroll, invoicing and secure RBAC across dev, stage and prod.

    Show details
    • CRM Platform. Engineered a full-scale CRM covering payroll, timesheets, placements, invoicing, onboarding, and document workflows with React, Material UI, Redux, and Firebase.
    • Performance. Built 40+ reusable React components and cut UI load times by 30% via memoization and selector-based state management.
    • Backend & RBAC. Developed payroll, margin, and tax logic with Node.js, Express, and Firebase Cloud Functions; implemented secure RBAC with Firebase Auth and Firestore rules.
    • Quality & CI/CD. Resolved 150+ production defects and enhanced CI/CD with Firebase Hosting and GitHub Actions across dev/stage/prod.
Fig. 06 · About

About

Full Stack AI Engineer with 6+ years building production web, mobile, and AI systems. I've spent most of that time on one HIPAA-compliant clinical platform at Atelia, growing from a contract developer in India to lead engineer in the US. Today I own the platform end to end: 24 backend services, 9 web apps, and 3 mobile apps used daily by doctors, hospitals, and patients.

Most recently I've built the platform's AI layer: a live ambient scribe that drafts clinical notes during doctor-patient visits, and 30+ AI agents that handle documentation, summaries, and clinical workflows, with clinicians reviewing anything before it reaches the record. I care about making AI reliable enough for high-stakes work, not just impressive in a demo.

I also own delivery: I led a migration that cut infrastructure costs by about half, and I'm the team's primary code reviewer (840+ pull requests). Outside work I design and ship products solo, a few of which are above.

Stack

Frontend & Mobile
React, Next.js, React Native, TypeScript, Redux, Zustand, Tailwind CSS
Backend & APIs
Node.js, NestJS, Express.js, PostgreSQL, Firestore, Prisma, REST APIs
Cloud & DevOps
Docker, Kubernetes, Serverless, Terraform / OpenTofu, GitHub Actions, CI/CD
Cloud Platforms
Google Cloud, AWS, Firebase, Vercel
AI & Real-time
LLMs, RAG, pgvector, Embeddings, WebSockets, Twilio, HIPAA

Education and recognition

  • MS in Computer Science

    Kennesaw State University. GPA 3.83/4.

  • BTech in Computer Science

    Sasi Institute of Technology and Engineering. GPA 9.59/10, top 1% of graduating class.

  • IEEE Xtreme

    24-hour programming competition. All India Rank 62 among 4000+ participants.

  • HackerRank certificates

    Node.js, React, SQL, REST API. Certified across multiple domains.

  • Open source

    50+ GitHub repositories. TypeScript and Node.js.

Fig. 07 · Contact

Let's talk

Hiring for a Full Stack or AI engineering role? Send a note here, or ask the assistant to pass one along.

Resume