Role: AI Engineer
AI-Native Product Engineering | WFO Bangalore | 4+ years
About TPH
The Product Highway is an AI-native product strategy and engineering firm that partners with businesses from the earliest spark of an idea all the way through to enterprise scale. We grew from zero to multi-million ARR within our first year, working with clients across India, APAC, Europe, and North America.
The conventional product/software industry is coming to an end. The line between product development and business development is disappearing. The best talent won't spend a career maintaining one product. They'll operate as forward-deployed product managers and engineers, building businesses end to end in small, high-leverage teams, one after another. That's the vision we're building toward.
AI got very good at the how: shipping code faster than ever. But almost nothing has changed in the what: deciding what to build, why, and in what order. That's where we live. We believe software is an approximation of the real world, and what matters isn't lines of code or sprint velocity. It's whether the solution actually maps to how a business works, how customers think, and how value gets created.
From in-house AI project managers to near-universal adoption of tools like Cursor and Claude, we've rethought every conventional process from first principles. AI isn't a feature we offer. It's how we think, build, and deliver. The result: legacy systems rebuilt in under 6 months, enterprise platforms built in 3, mobile applications shipped in under 1.
We only take on problems we find genuinely interesting and worthy of being solved, and we refuse to ship anything we wouldn't stand behind. Every hire gets us closer to that standard.
Who we're looking for
Regardless of role, every person at TPH shares these traits:
- End-to-end owners. You own outcomes, not tasks. If something you're responsible for falls through a crack, that's on you. Not because someone assigned it, but because you wouldn't have it any other way.
- Clear, direct communicators. Bad news doesn't get better with age. Context doesn't transfer through vague Slack messages. You surface problems early and communicate with precision.
- Uncompromising on quality. You have a quality bar that's yours, not your manager's. You won't ship something you wouldn't stand behind, even under deadline pressure.
- AI-obsessed. You see AI as how work gets done, not a nice-to-have. If you're still doing something manually that AI could handle, you feel that as friction, not normalcy.
- Structured thinkers. When faced with an ambiguous problem, you break it down, reason through the trade-offs, and arrive at a position. You don't wait for someone to tell you the answer.
- Experienced enough to use AI systematically. You have enough depth in your craft that AI makes you dangerous, not dependent. You direct it, evaluate its output, and know when it's wrong.
Why we're hiring
AI systems are now the core of most of what we ship: voice agents, sales copilots, moderation pipelines, analytics agents, retrieval-heavy products. We need an engineer who has actually taken AI systems to production, kept them there, and knows the difference between a demo and a system that survives real users.
The work
You'll build AI systems across TPH's client portfolio: conversational agents for e-commerce, multi-tenant agentic architectures for sales intelligence, voice and audio pipelines for gaming moderation, and internal platforms for agent evaluation and benchmarking. The projects change. The bar doesn't: production quality, measured behavior, no vibes-based shipping.
Stack exposure: Claude and Gemini APIs, open-weight models (Gemma, Sarvam), RAG pipelines with pgvector, agent frameworks and custom orchestration, STT/TTS pipelines, eval harnesses, Python/TypeScript, Postgres, AWS.
What you'll achieve
- Ship AI systems that real businesses depend on, across multiple domains, not one product for years
- Build genuine depth in the parts of AI engineering that matter in production: retrieval quality, eval design, agent reliability, cost and latency budgets
- Work directly with founders and client stakeholders. Your technical judgment shapes what gets built, not just how
- Learn how model selection, architecture, and evals connect to business outcomes, because on every project the client is paying for outcomes
What you'll do
- Design and build RAG systems end to end: chunking and indexing strategy, retrieval quality measurement, grounding, failure handling
- Build agentic systems: intent routing, tool use, multi-step orchestration, guardrails, graceful degradation when the model is wrong
- Own evals as a first-class deliverable. Every AI feature ships with a way to measure whether it works, regression-test changes, and catch drift
- Make model selection calls with reasoning: when a frontier model is worth it, when a small open-weight model wins on cost and latency, when fine-tuning beats prompting
- Take systems to production and keep them healthy: observability, cost tracking, latency budgets, prompt and version management
- Work across the stack when needed. Fix the API, the queue, or the schema that's blocking you rather than filing a ticket
Must-haves
- 4+ years of overall engineering experience with strong software fundamentals. AI engineering on top of weak engineering doesn't work
- Built at least one AI system that is running in production with real users. You can walk us through what broke, what you measured, and what you'd do differently
- Hands-on depth across the standard AI stack: RAG, evals, and agents. Not tutorials, not POCs that died in a notebook
- Has written evals for an LLM system: defined what good looks like, built the harness, used it to make ship/no-ship decisions
- Strong Python; comfortable in TypeScript or willing to get there fast
- Can explain trade-offs clearly to non-AI stakeholders: why retrieval is failing, why the agent loops, what the fix costs
- Deep understanding of Docker, Kubernetes, CI/CD pipelines, and cloud infrastructure (AWS/Azure/GCP)
- Practical expertise implementing pgvector at scale (custom chunking, distance metrics, HNSW tuning) alongside hands-on experience using at least one dedicated enterprise vector database (Pinecone, Qdrant, Milvus, or Weaviate)
Strong pluses
- Voice/audio pipelines: STT, TTS, streaming, latency-sensitive real-time systems
- Experience with open-weight models: deployment, quantization, on-device inference
- Fine-tuning or post-training experience with a clear reason it beat prompting
- Built internal tooling for AI development: eval platforms, benchmarking harnesses, prompt management
- Multi-tenant AI architectures or per-client configuration systems
AI-native expectations
- Uses Claude Code or Cursor as the primary development environment, shipping production code through AI-assisted workflows daily
- Maintains CLAUDE.md and project context files so AI tools know your conventions, architecture, and constraints
- Plan-first for multi-file changes: AI reads the codebase, you approve the approach, then execute
- Uses AI to accelerate the AI work itself: generating eval cases, stress-testing prompts, reviewing agent traces
- Tests after every AI-generated change. AI writes fast; you keep it honest
How to apply
Send your resume and a short note on why this role excites you to anju@theproducthighway.com. Skip the generic cover letter—we'd rather hear about the AI systems you've built, the real-world problems you've solved, the models, frameworks, or tools you've worked with, and what you can bring to The Product Highway. If building AI-first products and working across cutting-edge applications sounds like the kind of challenge you're looking for, we'd love to hear from you.