SharjeelAnsar
Full-Stack AI Engineer. I build production AI systems end to end: voice agents and agentic AI on top, plus the frontend, backend and data layer they actually run on underneath.
- FrontendReact, Next.js
- BackendNode.js, NestJS, PostgreSQL
- AI integrationRAG, MCP, LLM pipelines
- Voice & agentic AIVapi, multi-agent systems
- 80%
- Cut in per-call cost for voice AI systems
- <100ms
- Real-time search latency, down from 10 seconds
- 50+
- Business locations running systems I built
- 3+
- Years shipping production AI systems
Voice agents that answer real calls and book real appointments
Businesses lose bookings to voicemail and hold queues. I build the voice agents that answer instead: live in production, wired into the systems a business already runs on, and tuned to keep cost per call down as volume grows.
150+
Calls handled per day
live in production
80%
Lower cost per call
after pipeline redesign
~55%
Smaller system prompt
faster live responses
MCP
Custom tool layer
scheduling + data sync
How a call flows
- 1
Patient call
Inbound voice
- 2
Vapi voice layer
Speech + orchestration
- 3
LLM + tuned prompt
Trimmed prompt, knowledge base
- 4
Custom MCP tools
Scheduling, data sync
- 5
Clinic EHR
Booked & recorded
Built from scratch, running live
Vapi, Next.js and Python, taken from an empty repo to a production system handling 150+ calls a day, architected to scale well beyond that.
Cut cost per call by 80%
Redesigned the voice AI pipeline architecture and call orchestration logic, cutting operating cost per call by 80%.
Connected to real clinic systems
A custom MCP tool layer wires the agent into EHR systems and third-party services, enabling automated scheduling and reliable data sync.
Tuned for live conversation
Targeted prompt engineering cut the system prompt by roughly 55%, speeding up responses mid-call; a structured knowledge base lifted conversational accuracy and drove client adoption.
Built to absorb volume
Python background workers process calls asynchronously, so sustained growth in call volume does not degrade service.
What I build for clients
Agentic and voice AI, plus the full-stack engineering underneath it. Hire me for one layer or the whole thing. Every item below is something I have shipped to production.
AI voice agents
Voice receptionists and automated call handling that book appointments, answer questions and hand off cleanly, built on Vapi and tuned for live conversation latency.
150+ calls/day in production for healthcare clinics
MCP tool layers & integrations
Give an agent real capabilities. I build custom MCP tool layers that connect LLMs to the systems you already run: EHRs, CRMs, schedulers and third-party APIs.
Custom MCP layer wired into clinic EHR systems
RAG & knowledge bases
Agents that answer from your documentation instead of guessing. Retrieval pipelines, vector embeddings and structured knowledge bases that measurably improve accuracy.
RAG agent systems at Objex; knowledge base at CCRIPT
LLM cost & latency optimisation
Already running an agent that is too slow or too expensive? I audit the prompt, pipeline and orchestration path and bring both numbers down.
80% lower cost per call; ~55% smaller system prompt
Multi-agent architectures
Coordinated agents that hand work between each other reliably, with message-based communication and distributed workflows that survive real traffic.
Pub/Sub multi-agent coordination at Objex
AI automation pipelines
The unglamorous automation that removes manual work: lead routing, notifications and data operations wired together with Zapier, Make and n8n.
Lead capture + Slack pipelines at Accident Payments
Full-stack product development
Whole products, not just the AI part. Next.js and React on the surface, Node.js and PostgreSQL underneath, shipped and maintained end to end.
5 CRM portals owned end to end at Accident Payments
Backend, APIs & real-time systems
REST and GraphQL APIs, microservices, WebSocket real-time sync and background workers, designed to stay fast as traffic grows.
50% faster response time; CRM search cut to under 100ms
Frontend engineering
Accessible, responsive interfaces in React and Next.js: dashboards, portals and product surfaces that hold up on every screen size.
KPI dashboards and CRM portals shipped to sales teams
Have an agent idea that needs to actually ship?
Available for freelance and contract work in Islamabad, Pakistan, working remotely.
Where I've worked
Agentic AI, full-stack products and the distributed backends underneath them.
CCRIPT Agency
May 2026 — PresentBuilt a production AI voice receptionist for healthcare clinics from scratch, and the MCP tool layer, background workers and prompt engineering that keep it fast and accurate.
- Built a production AI voice receptionist for healthcare clinics from scratch using Vapi, Next.js and Python, currently handling 150+ calls per day and architected to scale well beyond that.
- Cut per-call operating cost by 80% by redesigning the voice AI pipeline architecture and optimising call orchestration logic.
- Connected the voice agent to clinic EHR systems and third-party services by building a custom MCP tool layer, enabling automated scheduling and reliable data sync.
- Engineered Python-based background workers for asynchronous call processing, supporting sustained call volume growth without service degradation.
- Cut the voice agent's system prompt by roughly 55% through targeted prompt engineering, speeding up LLM response time during live patient calls and reducing load on the context window.
- Improved conversational accuracy and drove rapid clinic adoption by integrating a structured knowledge base and running iterative prompt engineering cycles.
Accident Payments
Dec 2025 — May 2026Owned five CRM portals end to end, from normalised PostgreSQL schemas and real-time sync to AI automation pipelines and power-dialer integrations.
- Owned end-to-end development of 5 CRM portals serving distinct sales and retention teams, architecting and maintaining each from the ground up.
- Designed normalised PostgreSQL schemas on Supabase and engineered real-time sync via Supabase Realtime, keeping state consistent across all 5 portals.
- Cut manual work in lead routing and client data operations by building AI automation pipelines with Zapier, Make and n8n, including a real-time lead capture and Slack notification system.
- Integrated Aloware and Aircall power dialers into the CRM portals for click-to-call and auto-dial workflows, and shipped KPI dashboards for team-level sales visibility.
Invitrex
May 2025 — Dec 2025Optimised the API architecture behind a restaurant management platform used by 50+ restaurants in Europe, and helped bring an ElevenLabs calling agent into production.
- Optimised API architecture using efficient data structures and algorithms, improving data retrieval and processing performance across a restaurant management platform used by 50+ restaurants in Europe.
- Contributed to an AI agent calling system built with ElevenLabs, bringing automated customer call handling into production.
- Built full-stack features for an e-commerce web application using React.js, focused on scalable frontend-backend integration.
Objex
Jan 2024 — Apr 2025Engineered RAG-capable AI agent systems and the TypeScript/NestJS microservices running them on GCP, plus the gateway layer routing traffic to them.
- Improved system response time 50% by developing and deploying TypeScript/NestJS microservices on GCP with algorithm optimisation.
- Engineered RAG-capable AI agent systems with multi-agent communication over Google Pub/Sub to coordinate distributed AI workflows.
- Reduced CRM search latency from 10 seconds to under 100ms by building real-time WebSocket connections for data retrieval.
- Managed microservice routing on AWS API Gateway and later migrated it to Apigee, improving traffic control, rate limiting and analytics.
Selected projects
Things I built end to end, from data ingestion and model integration through to the interface.
What I work with
Grouped by how I actually use them, not by how impressive the list looks.
AI / LLM
01Where most of my work lives
- AI Agents
- RAG
- Vector Embeddings
- MCP (Model Context Protocol)
- Multi-Agent Systems
- Prompt Engineering
- Knowledge Base Integration
- Vapi Voice AI
- Google Pub/Sub
Languages
02Daily drivers first
- TypeScript
- JavaScript
- Python
- Java
- C++
- HTML5
- CSS3
Frontend
03Product surfaces
- React.js
- Next.js
- Vite
- Hooks & function components
Backend
04Services, contracts, jobs
- Node.js
- Express.js
- NestJS
- RESTful APIs
- GraphQL
- Async programming
- Background workers
Databases
05State and storage
- PostgreSQL
- Supabase
- MongoDB
- Firestore
- SQL
- NoSQL
Cloud & DevOps
06Where it runs
- AWS (API Gateway)
- Google Cloud Platform
- Docker
- Kubernetes
- Vercel
- Firebase
- Apigee
- GitHub Actions
- CI/CD
- Microservices
Automation
07Removing manual work
- Zapier
- Make
- n8n
- Slack Integrations
- Aloware
- Aircall
Testing & Workflow
08How I work
- Jest
- Mocha
- React Testing Library
- Cypress
- TDD
- Agile/Scrum
- ClickUp
- Git
About
I own features from idea to production, not ticket by ticket.
I am a Full-Stack AI Engineer with 3+ years shipping production AI agent and LLM-powered systems end to end: RAG pipelines, MCP tool integrations, prompt engineering and multi-agent architectures, plus the Next.js, Node.js and PostgreSQL infrastructure they run on.
Right now I build agentic AI at CCRIPT Agency, where I took an AI voice receptionist for healthcare clinics from scratch to 150+ calls a day, cut per-call cost by 80%, and connected it to clinic EHR systems through a custom MCP tool layer.
Before that I owned five CRM portals end to end at Accident Payments, and engineered RAG-capable agent systems and microservices at Invitrex and Objex. That work took CRM search latency from 10 seconds to under 100ms.
Education
COMSATS Institute of Information and Technology
Sep 2020 — Aug 2024BS, Computer Science
Let's put an AI agent into production.
Available for freelance and contract work: voice agents, agentic AI and RAG systems, or the full-stack frontend and backend engineering around them. Happy to take one layer or the whole build.