Ishaan Bhalla
Full-Stack Engineer specializing in Voice AI and Agentic Systems — building technology for the problems that matter most
About Me
I'm a full-stack engineer who builds AI products from the ground up — from early architecture decisions to production systems handling thousands of users daily.
Right now, I specialize in voice AI and agentic systems. At Paratus Health (YC W25), I built the entire voice platform as the sole engineer — real-time pipelines, LLM orchestration, agent management tooling — scaling to 10K+ calls/day across 35+ clinics. Before that, I built AI mental health tools at Cognovi and internal frameworks and data pipelines at Albert Invent and Virtusa.
I've often been the first engineer on the team, which taught me to own problems end-to-end. But the work I'm most proud of has always involved close collaboration — with founders, designers, and end users. I build best when I'm close to the problem and the people it affects.
Education
San Jose State University, San Jose, CA
Rajiv Gandhi Technical University, India
Experience
Founding Engineer
Paratus Health (YC W25)
July 2025 - April 2026
- Joining as the sole engineer at a Y Combinator healthcare startup meant owning everything from day one. The mission? AI voice agents that make healthcare communication faster and more human.
- I architected the voice-first infrastructure using LiveKit (Python microservice), with Twilio SIP trunking and multiple voice providers (Retell AI, ElevenLabs, Cartesia) to achieve sub-500ms response times. The platform now handles ~10K patient calls daily across 35+ clinics — including a pediatric after-hours system where accuracy is non-negotiable because a child's health could be at stake.
- Designed an end-to-end agent iteration pipeline: client feedback captured via Granola, piped through Claude Code via MCP, with prompt versioning in Langfuse — enabling the non-technical team to create and refine voice agents without writing code. Cut new client onboarding from days to hours.
- Beyond voice, I built the full communication layer — Vonage SMS, SendGrid email, and Playwright RPA automations to push data into EHR systems. Every channel had to be reliable because every message matters.
- On the quality side, I used Hamming.ai to simulate thousands of test calls and built a client-facing feedback loop that feeds directly into prompt tuning through LangFuse. The LLM gets better with every call.
Founding Engineer
Cognovi
July 2024 - June 2025
- At Cognovi, I stepped in to tackle a big challenge: making mental health AI feel more trustworthy and insightful. We were building tools that detect cognitive distortions in user input — a powerful idea, but the outputs weren’t always consistent or reliable.
- I designed and deployed an agentic system that improved how the AI interprets and responds to emotional content, leading to more meaningful journaling experiences. I also built a custom evaluation framework to measure AI performance — not just in accuracy, but in emotional tone and helpfulness.
- On the product side, I led full-stack development (Next.js frontend, Express + OpenAI backend), and containerized the app for rapid deployment on AWS, cutting dev time in half.
Full Stack Data Engineer
Albert Invent
October 2023 - July 2024
- Albert was growing fast — with distributed teams shipping features rapidly, UI consistency became a bottleneck. Developers were duplicating effort, and product quality started to drift.
- I partnered with the CTO to build “Maya,” a shared frontend framework using HTML web components and vanilla JS, giving the team a clean, reusable system that cut dev time in half and brought design coherence across the app.
- Alongside that, I built backend services in Go and optimized data pipelines using Ray and S3, reducing processing errors and speeding up retrieval across Postgres and DynamoDB.
Full Stack Engineer
SWAY
September 2022 - October 2023
- When I joined SWAY, the product was in its early stages and needed clear technical direction. I worked closely with the CEO to shape the foundation — from designing the database schema to building modular components and crafting a cohesive frontend experience.
- Using React, TypeScript, and GraphQL, I optimized key flows that improved performance by 40% and significantly reduced query times. These upgrades played a big role in growing user engagement and satisfaction.
Data Engineer
Virtusa
May 2019 - August 2021
- Virtusa’s machine learning teams needed cleaner, faster data — but the pipelines were breaking under scale. I designed Spark-based ETL flows and Kafka queues that could handle 2x more load, then restructured their AWS stack to support real-time workflows.
- The result: faster data ingestion, more reliable model training, and happier analysts.
Projects
AI Voice Agent
Medha
AI Voice Agent
An AI-powered voice assistant that can perform tasks and answer questions about my experience.
AI Job Search Assistant
Seeker
AI Job Search Assistant
An agent-powered job search platform that streamlines job discovery and outreach using OpenAI.
AI Climate Coach
Footprint AI
AI Climate Action Coach
An AI-powered climate coach that tracks your carbon footprint and delivers personalized sustainability recommendations.
Skills & Technologies
Programming
- Python
- TypeScript / JavaScript
- Go
- C++
AI & Voice
- Claude API
- OpenAI API
- LangChain
- LangFuse
- LiveKit
- Retell AI
- ElevenLabs
- Cartesia
- Hamming.ai
AI Engineering
- LLM Orchestration
- Prompt Engineering
- LLM-as-Judge Evals
- Agent Design
- Voice Pipeline Architecture
Frontend
- React
- Next.js
- Tailwind CSS
- Material-UI
Backend
- Nest.js
- Node.js
- Express
- FastAPI
- GraphQL
- WebSockets
Communication & Integrations
- Twilio (SIP Trunking)
- Vonage (SMS)
- SendGrid (Email)
- Playwright (RPA/Automations)
Databases
- PostgreSQL
- DynamoDB
- MySQL
- Supabase
- MongoDB
DevOps & Cloud
- AWS (EC2, S3, Lambda)
- Docker
- Kubernetes
- Terraform
- GitHub Actions
- CI/CD