Senior Engineer — Applied AIAetos Design & Engineering Pvt. Ltd. · HSR Layout, Bengaluru · On-site

About AetosFor 22 years we have been an engineering company. Computational fluid dynamics, thermal and structural analysis, design, simulation and validation — delivered to clients in railways and metro, automotive, aerospace and space, defence, marine and shipbuilding, electric motors and drives, oil and gas, and power.

Our work is the unglamorous, consequential kind. Whether a metro car's saloon holds a uniform temperature across a summer afternoon. Whether a structure survives its duty cycle. Whether a design meets a standard that a certifying authority will hold it to. Our outputs go into approval chains, get signed, and get archived, and they have to be right.

Why we are building this teamEngineering is an information-heavy business, and ours has accumulated a great deal of it — analysis reports, specifications, test and trial data, standards, drawings, correspondence, and two decades of completed project archives across a dozen industries.

Very little of it is reachable by software. Engineers find what they need by knowing where to look, or by knowing who to ask, and as people move on, the second route closes. The same is true, at much larger scale, inside most of the organisations we work for.

At the same time, a good deal of routine engineering effort is mechanical rather than intellectual — preparing cases, extracting values, assembling reports, checking one document against another. That effort is real and it is expensive, and a meaningful part of it can now be automated in ways that were not practical three years ago.

We are starting an applied AI practice to work on both. You would be the first engineer in it.

The remitThe practice has a broad brief, and you will help decide where it goes first. The directions we see:

Engineering knowledge. Making the organisation's accumulated technical record — ours and our clients' — genuinely reachable, so that a question gets an answer rather than a filename.

Simulation and analysis support. Reducing the manual effort around solver work: case setup, parameter sweeps, post-processing, summarisation of results, and surrogate approaches where a full solve is unnecessary.

Technical documentation. Drafting, extracting,



comparing and checking the reports, specifications and compliance documents that engineering work produces in volume.

Client deployments. A large part of our client base operates under constraints that rule out public cloud entirely — regulated sectors, sensitive design data, security policies that do not permit data to leave their premises. Building systems that run well on infrastructure we do not own, and sometimes cannot reach, is a standing requirement rather than an exception.

Internal tooling. Estimation, quoting, project data, and the ordinary operational work of an engineering business.

You will not do all of this in year one. You will pick the two that matter most, do them properly, and build the team to take on the rest.

What the first year looks likeMonths 1–2. Understand the business — sit with the engineers, see how the work actually gets done, find where the effort goes. Choose the first problem with the team. Stand up the infrastructure: models running on our own hardware, the development environment, the deployment path.

Months 3–4. Build and ship the first system into real use. Not a demonstration — something engineers depend on. Alongside it, build the evaluation harness, because in this business an output nobody can check is an output nobody will use. Begin hiring: two engineers to start.

Months 5–7. Second workstream, first client-facing deployment, and a team that functions when you are not in the room.

The role grows into leading the practice. We are not looking for someone who wants to stay an individual contributor forever, and we are not looking for someone who wants to stop building.

What we needEssential

- 4–7 years writing production software in Python, to a standard you would defend in review
- You have taken an LLM-based system into production with real users — not a prototype, not a hackathon entry, not a notebook that worked once




- Hands-on with running models yourself: inference servers such as vLLM, SGLang or TGI, quantisation, and sizing hardware to a workload
- Production experience with vector search and retrieval — Qdrant, Milvus, pgvector, Elasticsearch or equivalent
- You have wrestled real-world data into usable shape: inconsistent formats, poor quality sources, things that were never designed to be machine-read
- Linux, Docker, and enough networking to diagnose a system you cannot reach from your laptop
- You can design an evaluation harness, and you instinctively reach for a measurement rather than an opinion

Strongly preferred

- Deployment into on-premise, restricted or otherwise constrained environments
- Clients or employers in regulated sectors — defence, government, banking, healthcare, heavy engineering
- Exposure to engineering, manufacturing or simulation as a domain, in any capacity
- Willing to travel within India and work at client sites
- You have mentored junior engineers and want to do more of it

Explicitly not required

- Model training, fine-tuning research, or publications. We are not building models. We are building systems around them, and the hard parts are elsewhere.
- Hyperscale experience. Our problems are about correctness under constraint, not requests per second.
- A degree from a particular institution. We will look at what you have built.

How we workSmall team, no layers, no product managers between you and the problem. You will talk to the engineers whose work you are changing, and to clients directly.

We are on-site in HSR Layout, five days a week. That is a deliberate choice for this role — you are the first hire in a new practice, you will be learning a domain by sitting next to people who know it, and later you will be training juniors who need someone in the room.

We prefer things that work to things that demonstrate. A system in daily use by six engineers is worth more here than an impressive slide.

Compensation and logistics₹12–15 LPA depending on experience, reviewed at six months. Negotiable for someone clearly beyond the band.

Hardware you need to do the work, including GPU access, is provided. Conference and training budget. Bengaluru, HSR Layout, in office.

📌 Senior Engineer- Applied AI (Bengaluru)
🏢 Aetos Design u0026 Engineering
📍 Bengaluru