Ishmeet Singh

OVERVIEW
Pluk is a Slack agent that turns your network into a hiring pipeline. Open a channel for a role and Pluk's agents search every teammate's first- and second-degree LinkedIn connections, score each person against the role, predict who is open to a move, and draft warm outreach or intro requests that send only when a human approves. Used by 20+ enterprise customers; backed by a16z.

ROLE
• CTO. Led the pivot from Jobreel into Pluk and built the product and the engineering team from scratch.

HIGHLIGHTS
• 19 LangGraph agents in production, coordinated through Slack channels and group DMs.
• Retrieval on Pinecone and Postgres vector search across Claude, OpenAI, and Gemini.
• Human-in-the-loop by design, with fair-hiring guardrails in every agent.
• Distributed architecture on GCP for asynchronous and background agent pipelines.

OVERVIEW
An AI phone receptionist for gyms, built into BurnOn AI. A caller dials the gym's number, and the agent answers with the gym's own classes, memberships, and prices, recognizes returning members and leads by their phone number, captures new leads, and writes a follow-up for staff after the call.

ROLE
• Built the agent end to end: the voice pipeline, its MCP server in the BurnOn API, and the post-call follow-up.

HIGHLIGHTS
• Real-time voice pipeline: Daily for telephony, Cartesia for speech in both directions, Gemini for the conversation, and Silero plus a turn-detection model for natural turn-taking, on Pipecat Cloud.
• A dedicated MCP server injects the call's business and caller number server-side, so the model can never look up another person's data.

OVERVIEW
Microsoft Teams is Microsoft's collaboration platform, used by over 250 million people a month. Ishmeet worked on the iOS client's Calling and Meetings experience, the real-time audio and video core of the app.

ROLE
• Built consumer features including audio-only mode, the Teams Premium green room, and screen share mode.
• Built the telemetry pipeline behind Calling and Meetings' most important reliability scenarios, which became a core input to leadership's OKRs, and helped raise key meeting scenarios to 99.5% reliability.
• Led a cross-org effort that cut the mobile core media stack's size by 10%.

CORE TECH STACK
• iOS, Swift, SwiftUI, Combine, RxSwift, MVVM, video and audio processing, telemetry, SQL, Azure Data Explorer, Azure DevOps, CI/CD pipelines

OVERVIEW
Ray1 is a hyper-spectral device that fits in your pocket and allows you to measure the invisible. Ray1 is a Bluetooth-connected medical device that provides thermal data to analyze and proactively identify injuries under the skin.

This device connects with an iPhone to transmit thermal images, processed to provide in-depth data analysis to help healthcare providers identify paramedic injuries.

ROLE
• Lead engineer on the project with a team of over six cross-functional members, including engineers, product managers, and QA.

ACCOMPLISHMENTS
• Built the framework, analysis algorithms, and user interface to analyze the thermal images. The user interface allowed the user to annotate and extract specific data points for further analysis.
• Created the central Bluetooth Low Energy (BLE) service class to connect and manage multiple Bluetooth devices simultaneously.

CORE FEATURES
• Offline-ready app
• Globalization and localization
• Push notifications
• Mobile-first

CORE TECH STACK
• iOS, Swift, SwiftUI, Combine, RxSwift, MVVM, Core Bluetooth, core data, image processing