AI Applications Engineer

Position : AI Applications Engineer

Experience : 4+Years

Timings : 3PM-12AM IST

Location : Remote

About the role

We’re building agentic AI apps for real business use—voice / chat agents that orchestrate workflows across CRMs / ERPs and internal tools. You’ll help us shipfeatures end-to- end : Django-based backends, real-time agent infra (LiveKit + Deepgram), and LLM integrations (API and self-hosted). Great for someone with solid Python fundamentals plus side projects in AI / agents.

What you’ll do

Build backend services in Python / Django (REST / WebSocket endpoints, auth, RBAC).

Implement agent flows that combine LiveKit (real-time media), Deepgram (ASR / TTS), and LLM APIs (OpenAI / Anthropic / Mistral / Azure).

Integrate tools and data : PostgreSQL , Redis , external SaaS (e.g., Salesforce / HubSpot / Zendesk) via APIs.

Add job orchestration C background work : Celery / RQ , scheduled tasks, retries.

Support self-hosted LLMs (Docker / Kubernetes) and vector search (e.g., pgvector , Qdrant , or FAISS ).

Deploy and operate services on Render.com (or similar) : blue-green deploys, background workers, cron jobs, service health checks, environment variables / secrets.

Write tests, logs, and metrics; add basic observability (Prometheus / OpenTelemetry, Sentry).

Document endpoints, flows, and runbooks.

Nice to have (you’ll learn here if you don’t know all yet)

LangChain / LlamaIndex , function / tool calling, multi-step “agents”.

Real-time comms basics : WebRTC , TURN / STUN, audio pipelines.

Prompt engineering, RAG patterns, evals C guardrails (prompt libraries, regex / JSON schema validation).

Frontend basics (React) to debug agent UIs.

CI / CD on Render.com (build C start commands, service dependencies, health checks, persistent disks).

Security hygiene : OAuth2, JWT, signed webhooks, PII handling, rate limits.

Minimum qualifications

~3 year experience (internships or personal projects count) with Python andDjango / FastAPI.

You’ve shipped at least one project using an LLM (API or local) or a speech API.

Comfortable with Git , Docker , PostgreSQL , Redis , and basic Linux .

Can read API docs and wire up third-party integrations quickly.

Clear written communication; habit of small, tested PRs.

Tech stack you’ll touch here

Python 3.11+, Django / DRF, FastAPI (selective), Celery, Redis, Postgres, WebSockets, LiveKit, Deepgram, LLMs (OpenAI / Anthropic / Mistral / Azure), self-hosted (Ollama / vLLM), vectorDB (pgvector / Qdrant),LangChain / LlamaIndex, Kafka (nice), Render.com, Docker, GitHubActions, Sentry / OTel.

Example problems you might work on

“Voice agent” that answers calls, verifies a customer, pulls order status from Django models, and schedules a return— LiveKit + Deepgram + tool-using LLM .

RAG : ingest PDFs / emails from a shared inbox, chunk / embed, store in pgvector, expose a search+answer endpoint.

Workflow : Celery pipeline to reconcile invoices nightly; retries with idempotency keys and alerting.