~/sreenivasan.ac
Portrait of Sreenivasan AC

Sreenivasan AC

AI Software Engineer, Bay Area

I build AI systems that hold up in production: agents, RAG pipelines and on-prem LLM deployments.

Nine-plus years across Google, Apple and high-growth startups, for enterprise, customer support, healthtech and legal teams.

Answers are matched in your browser from what's on this page. No LLM call, nothing is sent anywhere.

Selected projects

CallEnvoy landing page with a transcript of an AI phone call

CallEnvoy

Phone-call server for AI agents. Any MCP client (Claude Code, Codex, Claude Desktop, Grok) asks it to call a business; the AI presses the menu keys, waits on hold and talks to the person.

You can listen in and take over at any time. One Python process hosts the MCP endpoint, the telephony webhooks and a voice bot per call.

Voice agents, MCP, Pipecat, FastAPI, Twilio, Deepgram, GPT-4.1

Review Sieve audit comparing a business's listed ratings with its model-judged rating

Review Sieve

Search for a business and models read its reviews on Google Maps, Yelp, Tripadvisor and Trustpilot. You get an honest rating, the problems to watch for, and every review tagged genuine, generic, empty or suspicious.

A cheap model tags every review and an expensive one reads only the genuine ones, so a run stays low-cost. The cost is shown before it starts.

LLM pipeline, model cascade, Grok, web data extraction, Python

DeskRecall profiles window showing a saved window layout per display and desktop

DeskRecall

macOS menu bar app that puts every window back where it belongs for whichever displays are connected: the right display, the right Mission Control desktop, the right size, minimized or not.

Save a profile once per display setup; plugging a display in does the rest.

Swift, macOS, menu bar app

Dollhouse view of a to-scale 3D model of a one-bedroom apartment

Spruce Apartment 3D

Walkable, to-scale Three.js model of a one-bedroom apartment, built to plan furniture before moving in, with dollhouse, plan and first-person walk views and an iPhone AR page.

No external 3D assets or image textures: every surface is drawn procedurally on canvas.

Three.js, WebGL, procedural textures, AR

Sandhyāvandanam app home screen showing the evening sandhyā

Sandhyāvandanam

Web app that guides practitioners through Sandhyāvandanam: step-by-step actions, mantras in the script of your choice with transliteration, and meanings.

Covers the Kṛṣṇa Yajur Smārta and Śrī Vaiṣṇava traditions for all three daily sandhyās. In alpha.

React, TypeScript, Tailwind CSS, PWA

Brahmasumm document AI answering a question with citations

Brahmasumm Document AI

RAG platform with on-prem LLM deployment, multimodal ingestion, citations and other AI features for Fortune 500 teams.

Enterprise deployment: secure document intelligence for Lawrence Berkeley National Lab and ATMUS Filtration.

Azure, LangChain, vector DBs

File Vyasa proposing a folder structure

File organiser AI agent

Desktop assistant that understands the files on your machine, proposes folder hierarchies, and runs safe move and rename operations with human oversight.

Privacy-first architecture with Tauri, Agno agents, and safe execution planning.

Agentic workflow, document understanding, local LLM, Agno

AI Receptionist dashboard

AI Receptionist (Keystone)

AI receptionist chatbot for self-care businesses such as salons, medspas and fitness studios, with booking, CRM, and marketing automation.

Full-stack SaaS with an embeddable chat widget, admin dashboard, appointment scheduling, lead capture, SMS campaigns, and analytics.

Agentic AI, Agno SDK, FastAPI, React, OpenAI

Constella clustering visualization

Constella

Token-efficient auto-grouping and auto-labelling pipeline for large text collections like bookmarks.

Combines provider-agnostic embeddings, deterministic K-Means, and UMAP visualisation for reproducible grouping and labelling at scale.

LiteLLM, deterministic K-Means, UMAP

Experience

Keystone Founding Engineer, AI Agents for Sales and Marketing One of the first engineers. Built AI agents that run sales and marketing for brick-and-mortar businesses, including a voice agent for missed calls. Jan 2026 – Dec 2026 Bangalore, India
  • One of the first engineers at this early-stage startup, whose AI agents run websites, social media, lead follow-up and CRM for salons, medspas, fitness studios and other brick-and-mortar businesses. Built a major part of the codebase from scratch and made early decisions on the agentic architecture, engineering practices and hiring.
  • Designed and built Front Desk, a voice agent that answers a business's missed and after-hours calls: it answers from the business's own information, texts a booking link, takes messages and creates a lead for every call, with summaries, transcripts and recordings in the conversation thread.
  • Designed the voice agent's evals: a simulated-caller regression suite run before each prompt release, and LLM-judge scoring of live calls.
  • Built social posting end to end: Facebook and Instagram OAuth, AI-generated posts, scheduled publishing, carousels, and engagement sync for likes, shares and views.
  • Built lead and contact management (CRM) across backend and frontend: website lead forms, duplicate-contact review, hot-lead classification, notification emails and Meta Conversions API tracking.
  • Built LLM-based business data import that crawls a business's existing website, PDFs and images, with an operator review step that selectively merges the extracted fields.
  • Built the company design system, generating components from Figma designs with AI coding workflows.
  • Implemented the blog feature with AI post generation and an auto-scheduling agent.
  • Improved production reliability: cloud monitoring alerts and a production dashboard, AI code review on every pull request, and integration tests cut from 15 minutes to 3.

AI agents, voice agents, evals, Python, TypeScript, Google Cloud, Kubernetes, Meta APIs, Crawl4AI, design systems

Brahmasumm AI Founding Engineer, Enterprise AI Document Intelligence RAG pipelines over 50GB+ datasets and on-prem LLM deployments for Fortune 500 clients. Nov 2024 – Jul 2025 Bangalore, India
  • Designed and developed RAG pipelines processing 50GB+ datasets for Fortune 500 clients, implementing multi-modal extraction for PDFs, presentations, and spreadsheets.
  • Successfully deployed secure on-premises LLM solutions for clients Lawrence Berkeley National Lab (discoverers of 18+ periodic elements) and NYSE-listed ATMUS Filtration.
  • Achieved sub-second query response times through intelligent chunking strategies and vector database optimization (ChromaDB, Pinecone).
  • Spearheaded end-to-end development of critical features including multi-modal file processing, citation functionality, and knowledge base partitioning.

Enterprise RAG, agentic systems, Azure Document Intelligence, vector DBs, on-prem LLMs

Google Senior Software Development Engineer, Google Cloud Platform Planning and task-generation system for Google Cloud datacenters, handling 1000+ daily operations. Mar 2020 – Jun 2022 Sunnyvale, California
  • Developed end-to-end implementation of planning and task generation system for Google Cloud datacenter operations.
  • Created dynamic planning system to transform static Excel-based workflows into dynamic, automated processes handling 1000+ daily operations.
  • Engineered automated workflow for 200+ rack migrations in Google datacenters, reducing manual intervention by 80%.
  • Built intelligent reachability zone calculation system for server configurations, optimizing fiber utilization and reducing infrastructure costs by 15%.
  • Implemented comprehensive alerting framework for build workflows, reducing mean time to detection (MTTD) from hours to minutes.

Python, Kubernetes, workflow automation, network ops

Apple Senior Software Development Engineer (Contract), Cloud Technologies Cloud Usage Insights platform that enabled $240,000 in cost savings. Jun 2022 – Apr 2023 Sunnyvale, California
  • Designed and developed Cloud Usage Insights platform using Python microservices and RESTful APIs, delivering AI-driven analytics that enabled $240,000 in cost savings.
  • Developed data pipeline using Python cronjobs, aggregating cloud usage metrics from multiple sources.
  • Applied Software design patterns (Factory, Strategy) extensively for code maintainability and extendability of enterprise systems.

Python, REST APIs, cron jobs, cloud analytics, cost optimization

Fathom Health Software Development Engineer, ML Platform for Medical Coding OCR pipeline for 900K+ HIPAA-compliant medical records. Sep 2019 – Mar 2020 San Francisco, California
  • Architected OCR pipeline processing 900K+ HIPAA-compliant medical records, implementing text extraction and parsing for downstream Spark/BigQuery ML training.
  • Designed and deployed EFK (Elasticsearch, Fluentd, Kibana) stack for comprehensive log aggregation and analysis, improving system visibility and reducing troubleshooting time by 60%.

Healthcare AI, OCR, Apache Airflow, HIPAA

Invento Robotics AI Engineer, Humanoid Robot Platform Offline NLU and conversation system for the humanoid robot Mitra. Apr 2018 – Jul 2018 Bangalore, India
  • Developed offline NLU and conversation system for humanoid robot Mitra using RASA framework on NVIDIA Jetson TX2.
  • Implemented Alexa Skills using AWS Lambda for voice-controlled guest registration and recommendation systems.
  • Built chatbot interface using React.js/React Native with modular, extensible components.

Conversational AI, RASA NLU, NVIDIA Jetson, AWS Lambda

Groupon Software Development Engineer ReactJS deal-creation app for 500+ sales reps, and DocuSign automation saving $40,000+ a month. Jun 2014 – Jun 2017 Palo Alto, California
  • Co-architected mobile-responsive web application in ReactJS for Card Linked Offers, used by 500+ sales representatives to create Groupon deals.
  • Automated contract-signing process through DocuSign API integration, saving $40,000+ monthly.
  • Owned Contract Service and Fine Print API microservices, transforming static Google Spreadsheet workflows into dynamic API-driven processes.
  • Enhanced Deal Creation Tool with new features and performance improvements, supporting thousands of daily transactions.

ReactJS, Ruby on Rails, microservices

Skills

Languages
Python, TypeScript/JavaScript, Node.js, SQL, Bash scripting
AI / ML concepts
RAG, multi-agent systems, prompt engineering, semantic search, NLP, ML system design, LLM fine-tuning, observability
AI / ML frameworks
LangChain, LangGraph, Vercel SDK, OpenAI APIs and SDKs, vector DBs
Full stack
FastAPI, Flask, Express.js, REST APIs, React.js, Next.js, Tailwind CSS, Firebase
Cloud and DevOps
AWS, Google Cloud, Azure, Kubernetes, Terraform, CI/CD automation, Git
Database and pipeline
PostgreSQL, MongoDB, Drizzle, Prisma, Apache Airflow, Kafka
Web technologies
Selenium, Chrome extensions, OAuth2, RESTful APIs
Soft skills
Writing, structured thinking, problem-solving, stakeholder communication, curiosity

Education

Georgia Institute of Technology MS Computer Science, AI / Machine Learning Deep learning, computer vision, natural language processing, reinforcement learning. Graduate teaching assistant for Deep Learning (CS7643).
Aug 2017 – May 2019 Atlanta, USA
National Institute of Technology, Tiruchirappalli B.Tech Computer Science & Engineering ACM ICPC regional finalist, ranked #30 of 300+ teams (2013). All India Rank 4,120 among 2M+ candidates in IIT-JEE/AIEEE, the top 0.2%.
Aug 2010 – Jun 2014 Tiruchirappalli, India

Get in touch

I'm open to AI Engineer roles at AI startups and mid-stage companies in the Bay Area.

Email me Book a 15-min call Message on LinkedIn

Email
sreenivasan.nitt@gmail.com
Phone
+1 (404) 663-9270
GitHub
github.com/sreenivasanac
LinkedIn
linkedin.com/in/sreenivasan-ac
X
@sreenivasan_ac

I reply within 24 hours, and I'm happy to talk multi-agent systems, RAG pipelines and enterprise deployments.