AI Engineer (3+ years mandatory)

About Us
Businessonbot is a fast-growing B2B SaaS platform doing $1.5M ARR, serving D2C brands across India for WhatsApp-first marketing, support automation, and AI-powered customer journeys. Our stack connects WhatsApp, Instagram, Messenger, Email, and Voice channels to e-commerce platforms (Shopify, WooCommerce), payments, and CRM — all in real-time, at scale, for hundreds of active business clients.
We move fast, operate lean, and take reliability seriously because downtime for our clients means lost revenue

Role
We're looking for an AI Engineer who can own our agent stack end to end — design, build, ship, and keep it healthy in production.

This is not a research role. You won't be reading papers and running benchmarks in a notebook. Our AI agents talk to real customers of real D2C brands, on WhatsApp and on voice, about their orders and their money. When an agent hallucinates a discount code or loops on a tool call, a business owner notices and a customer is unhappy. So the job is: build agents that work reliably, prove they work with evals and traces, keep them from doing dumb or dangerous things with guardrails, and do all of it without the inference bill quietly doubling every quarter.

You'll be the person who holds the entire agent architecture in their head. That means being independent — you'll get direction on what the product needs, not a spec for how to build it

What You'll Own

1. Agent Architecture, End to End
2. Own the design and implementation of production LLM agents — tool calling, memory, state, multi-step reasoning, handoff to humans
3. Build and maintain Python/FastAPI services that expose these agents to our platform, reliably and at low latency
4. Decide agent boundaries: what's one agent vs. many, what's a tool vs. a workflow step, what should never be an LLM call at all

5. Integrate with our core backend, messaging channels, and e-commerce/CRM data sources so agents actually have the context they need

6. RAG & Retrieval Quality

7. Build and own retrieval pipelines — chunking, embedding, indexing, reranking, refresh strategy

8. Debug retrieval failures properly: is it the chunking, the query, the index, or the prompt? Know how to tell the difference

9. Handle multi-tenant data isolation in retrieval as a correctness and security requirement, not an afterthought

10. Keep retrieval fresh as brand catalogs, policies, and offers change

11. Guardrails & Safety

12. Design input and output guardrails — prompt injection defense, PII handling, scope enforcement, refusal behavior, escalation paths

13. Prevent the failure modes that cost customers money: hallucinated commitments, wrong order info, runaway tool loops, unbounded retries

14. Build deterministic fallbacks for when the model is wrong, slow, or unavailable

15. Treat "the agent said something it shouldn't have" as a production incident, not a prompt-tuning ticket

16. Observability, Tracing & Evaluation

17. Instrument everything — traces, spans, token usage, latency, tool call outcomes (we use LangSmith, New Relic, Sentry, Pino structured logging)

18. Build eval sets and regression suites so we know whether a prompt or model change made things better or worse, with evidence

19. Make agent behavior debuggable by someone who isn't you — a trace should tell the story without a walkthrough

20. Set up alerting for quality degradation, not just uptime

21. Cost Consciousness

22. Own the LLM and infrastructure cost of everything you ship (AWS, OpenAI, Gemini, vector store, embedding compute)

23. Know the unit economics: cost per conversation, per resolution, per brand — and be able to quote them

24. Make the routing calls — smaller model vs. larger, cached vs. fresh, retrieval vs. long context — and defend them with numbers

25. Flag cost anomalies early and propose architectural fixes, not just usage caps

26. Customer Empathy

27. Understand that our agents are the front door for D2C businesses talking to their customers

28. Read real conversations. Sit with the support and product teams. Let actual failures shape the roadmap, not intuition

29. Treat a customer-facing quality bug with more urgency than an internal refactor

30. Balancing Tradeoffs

31. Navigate the tension between shipping a working agent this month and building the platform that supports fifty of them

32. Know when a prompt is the right answer and when it's a hack that will break in three weeks

33. Make architecture decisions that hold up as the AI surface area grows — agents, voice, deeper e-commerce automation

What We're Looking For **
**Must Have

• 3–5 years of software engineering experience, with meaningful time spent shipping LLM-powered features to production
• Strong Python, with production experience in FastAPI (or equivalent async Python service frameworks)
• Hands-on experience building agents with LangChain (or LangGraph / similar orchestration frameworks) — not just calling a chat completion endpoint
• Practical experience with LangSmith or comparable tooling for tracing, evals, and prompt/version management
• Direct experience with OpenAI and Gemini model families — including their tool calling, structured output, and failure quirks
• Built and operated a RAG pipeline in production, and can explain what broke and how you fixed it
• Real experience designing guardrails and safety layers around LLM output
• Comfort with AWS and with reasoning about cost/performance tradeoffs in cloud and inference spend
• The ability to hold an entire agent architecture independently — you can go from a fuzzy product goal to a shipped, monitored system without hand-holding

Strong Plus

• Experience building voice AI pipelines — STT/TTS, streaming, turn-taking, barge-in, latency budgets
• Exposure to real-time messaging systems (WebSockets, pub/sub, queue-driven architectures)
• Experience with multi-tenant SaaS architecture and per-tenant data isolation
• Familiarity with WhatsApp Business API, omnichannel messaging, or e-commerce integrations
• Experience working alongside a TypeScript/Node.js backend
• Fine-tuning, distillation, or model routing work done for cost or latency reasons
• Prior experience at a startup in the 10–100 person range where you set the standard rather than inherited it

What We Don't Need

• Someone who needs a perfectly defined problem before they can start
• A prompt engineer with no software engineering depth — you will be writing and owning real services
• Someone who ships agents they can't observe, evaluate, or explain
• Someone who treats model cost as somebody else's problem
• Someone who optimises for their own comfort over team outcomes

How You Think

• You ask "how will this fail with a real customer?" before "how does this demo?"
• You don't trust an agent you can't trace
• You're bothered by silent failures, swallowed exceptions, and fire-and-forget async calls
• You reach for the smallest thing that works — a rule, a regex, a cached answer — before reaching for a bigger model
• You have opinions on architecture but hold them loosely when presented with better evidence
• You believe an eval set is more useful than an opinion about whether the new prompt is better
• You understand that a cost spike is an engineering problem, not just a finance problem
• You think about the engineer who has to debug your agent six months from now

What You'll Walk Into
We'll be transparent with you:

• A production AI surface that's growing fast — agents, voice workflows, and deeper e-commerce automation
• Real usage from hundreds of business clients, which means real edge cases and real feedback loops
• Genuinely good architecture in places, and real technical debt in others
• Evaluation and observability practices that exist but need to be made rigorous — you'll shape them
• A company that takes reliability seriously but is still building the culture around it
• This is a role for someone who thrives in that kind of environment — not someone who needs it to be clean before they can contribute.

What We Offer

• Meaningful ownership — you'll shape how engineering works here, not just execute someone else's vision
• Direct access to founders and product decisions
• Competitive compensation (cash + equity discussion based on profile)
• A product that real businesses depend on every day — your work matters immediately