AI & Agentic Systems
Production RAG pipelines and multi-agent workflows that survive real traffic — LangChain, LangGraph, LlamaIndex, CrewAI, AutoGen and MCP across OpenAI, Claude and Gemini, with vector search on Pinecone, ChromaDB or pgvector.
// available for global remote contracts — 2026
AI Technical Architect
& Full‑Stack Developer
for hire.
I'm Yashdeep Tomer — 15+ years of software engineering, now focused on what actually ships: production Voice AI agents, RAG systems and agentic workflows, backed by full-stack and DevOps depth. I design it, build it, deploy it, and stay until it works.
- 15+
- years engineering
- 33%→<5%
- missed-call rate, voice AI
- <2s
- voice agent turn-taking
- $0
- per-token cost, local LLMs
01 / services
What I build for clients
Voice AI Agents
Real-time voice agents that answer, qualify and route calls 24/7 — custom LiveKit STT→LLM→TTS pipelines, ElevenLabs Conversational AI, Deepgram, Twilio and SIP telephony. Sub-2-second turn-taking, measured.
Full-Stack Web & Mobile
End-to-end product engineering — Node.js, NestJS, Python and FastAPI backends; React, Next.js and Angular frontends; REST, GraphQL and WebSocket APIs designed for scale from day one.
DevOps & Cloud
AWS and GCP architecture, Docker and Kubernetes, Terraform IaC, and CI/CD pipelines that deploy with zero downtime — plus the observability (Grafana, Sentry, CloudWatch) to prove it's healthy.
02 / selected work
Recent results, not just résumé lines
-
24/7 Real-Estate Voice AI Platform
Architected the async backend powering an AI voice agent for inbound and outbound real-estate calls — caller identification, lead capture, agent routing and live warm transfers, bridging SIP/PSTN telephony to conversational AI over WebSockets.
-
Multi-Brand AI Lead Qualification — Finance Brokerage (Australia)
Three-brand voice agent covering inbound, outbound, after-hours and missed-call recovery. Missed-call rate dropped from 33% to under 5%; speed-to-lead went from next-morning to under 2 minutes.
-
Local-First Voice Pipeline — Zero Per-Token Cost
Custom STT→LLM→TTS stack running entirely on local models for on-prem privacy: ~0.5s transcription, ~300ms speech synthesis, sub-2-second end-to-end turn-taking — with no inference bill at all.
-
Enterprise Platforms — SpiceJet Cargo & Banking
Led architecture and cross-functional delivery for SpiceJet's Kargo360 logistics platform and a Saudi Arabian bank — Terraform-provisioned AWS infrastructure, EKS clusters and zero-downtime CI/CD.
03 / why hire me
Senior judgment, hands-on delivery
Architect who still codes
15+ years from IndiaMart and Monster.com to leading AI delivery — I design the system and write the production code myself.
Outcomes over demos
Every AI project ships with evals, observability and measurable business numbers — not a notebook that worked once.
US / EU friendly
Daily time-zone overlap, fluent written and spoken English, weekly demos, and async updates you can forward to stakeholders.
You own everything
Code in your repos, infra in your cloud accounts, NDA signed before we talk details. No lock-in, ever.
04 / stack
Technologies I work with daily
AI / GenAI
- OpenAI
- Claude
- Gemini
- LangChain
- LangGraph
- LlamaIndex
- CrewAI
- AutoGen
- MCP
- RAG
- Ollama
Voice AI
- LiveKit
- Whisper
- ElevenLabs
- Deepgram
- Twilio
- WebRTC
- Edge TTS
Backend & Data
- Node.js
- NestJS
- Python
- FastAPI
- GraphQL
- PostgreSQL
- MongoDB
- Redis
- Pinecone
- ChromaDB
- pgvector
Frontend
- React
- Next.js
- Angular
- TypeScript
- Tailwind
DevOps & Cloud
- AWS
- GCP
- Docker
- Kubernetes
- Terraform
- GitHub Actions
- Jenkins
- Nginx
05 / how we'll work
From first call to production
-
01
Discovery call — free
30 minutes on your goals, constraints and current stack. You leave with an honest read on feasibility, even if we don't work together.
-
02
Proposal & fixed estimate
Scope, architecture, milestones, timeline and price in writing within a few days. No open-ended hourly black holes.
-
03
Build with weekly demos
Working software every week in your repos and your cloud, with async written updates between demos.
-
04
Launch & support
Production deployment with monitoring and a warranty period — plus optional retainers for ongoing improvements.
06 / faq
Questions clients ask
What does it cost to build an AI voice agent or RAG system?
Most projects land between a 2-week proof of concept and an 8–12 week production build. After a free discovery call I provide a fixed-price proposal with milestones, so you know the full cost before we start.
Do you work with clients in US and European time zones?
Yes. I work remotely from India with daily overlap for US and EU hours, async updates in writing, and weekly demo calls. I've delivered projects for clients in Australia, the Middle East and India.
What engagement models do you offer?
Fixed-price projects with milestones, monthly retainers for ongoing development, and technical advisory for teams adopting AI. A free 30-minute discovery call is the starting point for all three.
Will you sign an NDA, and who owns the code?
Yes, I sign NDAs before any technical discussion. You own 100% of the code, infrastructure and IP — everything lives in your repositories and cloud accounts from day one.
Do you provide support after launch?
Every project includes a warranty period for fixes, and I offer ongoing maintenance retainers covering monitoring, model updates, cost optimization and new features.
Can you work with my existing team or codebase?
Yes. I've led cross-functional teams for 15+ years and regularly join existing codebases — audits, architecture reviews, AI feature additions, or leading delivery end to end.
07 / contact
Have a project in mind?
Let's talk.
Tell me what you're trying to build. I reply within one business day — usually faster.