Software Development Engineer — Agentic AI

About EloElo

EloElo is a consumer AI and entertainment company building the next generation of AI-native experiences. Our products span AI companionship and interactive entertainment — serving millions of users across India through real-time voice, personalised LLM pipelines, and emotionally intelligent AI characters.

We're now building the intelligence layer that makes our AI truly autonomous, context-aware, and deeply personal — and we need an engineer who thinks in agents, not prompts.

The Role

As SDE — Agentic AI, you'll design and ship autonomous AI systems that power real user experiences across two distinct product domains. You'll own everything from memory architecture to multi-agent orchestration — reasoning deeply about why agents fail and building systems that don't.

What You'll Build

You'll own the full spectrum of agentic experiences — from intimate one-on-one companion interactions to large-scale interactive entertainment and learning flows:

Persistent Memory Agents — long-term recall of user preferences, emotional milestones, relationship history, and behavioural patterns across sessions using layered vector + episodic memory stores

Emotional Context Synthesis — agents that continuously model a user's emotional state from conversation signals and adapt tone, pacing, and depth of engagement in real time

Proactive Engagement Agents — systems that detect disengagement signals, relationship drift, or mood shifts and autonomously initiate contextually appropriate interactions

User Intent & Goal Tracking — multi-session agents that track evolving user goals (wellness, learning, entertainment) and surface timely nudges or check-ins

Persona Consistency Engine — retrieval-augmented pipelines that keep AI character personality, memory, and voice stylistically coherent across thousands of concurrent users

Knowledge Synthesis Agents — systems that dynamically assemble and compress information from multiple sources into coherent, engaging narratives tailored to a user's knowledge level

Interactive Learning Agents — conversational agents that guide users through structured learning paths, adapt difficulty in real time based on comprehension signals, and proactively fill knowledge gaps

Tonality Synthesis — agents that modulate emotional register, energy level, and cultural flavour of generated content — from playful and casual to authoritative and instructive

Learning Retention Agents — spaced repetition and active recall systems embedded in entertainment flows, so users absorb knowledge without feeling like they're studying

Live Interaction Agents — event-driven agents that respond to real-time signals (live sessions, audience reactions, quiz events) with contextually intelligent interventions

What We're Looking For

Must-Have

2–5 years of hands-on software engineering experience

Deep fluency in agentic AI concepts — ReAct loops, tool use, planning, self-reflection, multi-agent orchestration, agent failure modes

Strong grasp of memory architecture — short-term buffers, long-term vector stores, episodic vs. semantic vs. procedural memory, and hybrid retrieval strategies

Production experience with vector embeddings — semantic search, dense + sparse hybrid retrieval, re-ranking, and embedding evaluation

Hands-on with at least one agent framework — LangGraph, CrewAI, AutoGen, or equivalent — and the depth to go beyond them when needed

Solid Python and comfort with frontier LLM APIs (OpenAI, Anthropic, Gemini, or open-source)

Experience building RAG pipelines — chunking, metadata filtering, query transformation, eval-driven iteration

Good to Have

Experience with multilingual NLP — especially Hindi / Hinglish language understanding and generation

Familiarity with voice AI pipelines (STT → LLM → TTS) and low-latency agentic integration

Prior work in consumer AI with real user traffic at scale

Exposure to personalisation systems — user embeddings, preference modelling, collaborative filtering

Knowledge of graph databases (Neo4j, Memgraph) for relationship-aware memory or knowledge graphs

Tech Stack

Python / Java / Go · LangGraph / CrewAI · Qdrant / Weaviate / Pinecone · OpenAI / Gemini / Anthropic APIs · FastAPI / Spring Boot / Gin · Redis · PostgreSQL · GCP / AWS · Docker / Kubernetes

What Makes This Role Different

Your agents will be in the hands of real users daily — not demos, not internal tooling

You'll work at the intersection of memory, voice, and personality — one of the rarest and most challenging surfaces in applied AI right now

Direct collaboration with the founding team with real architecture ownership

Fast feedback loops — your code ships quickly, and you see its impact in product metrics

What We Offer

Competitive compensation + ESOPs

Full ownership of your problem space

Access to frontier model APIs, GPU infra, and a team pushing the state of the art

A front-row seat to building AI products for the next billion users

EloElo is an equal opportunity employer. We believe great AI is built by teams as diverse as the users they serve.