Applied AI / Machine Learning Engineer

Location: Bhopal, Madhya Pradesh
About Lear Labs
Lear Labs is an enterprise AI lab that builds production AI systems for organizations across India and beyond. We are headquartered in Bangalore, with projects and partners spanning across the world.
We work on problems where AI has to operate in the real world, not just inside a demo. Our work spans computer vision, machine learning, agentic AI, large language models, data systems, automation, and enterprise software.
Our partners include large enterprises, infrastructure organizations, conglomerates, and government bodies. We take projects from raw data and early experimentation all the way to systems that people use operationally.
About the Role
We are looking for an Applied AI / Machine Learning Engineer based in Bhopal to work closely with our engineering and operations teams on one of our major infrastructure AI projects.
A key part of the work involves a road and infrastructure intelligence platform that processes large volumes of imagery and video collected from vehicles in the field. The system identifies road conditions and infrastructure assets, converts them into structured and geo-referenced data, and produces dashboards and reports used by operational teams and decision-makers.
However, this is not exclusively a computer vision role.
Depending on your strengths, you may also work on:
LLM and agentic AI systems
AI-powered data processing and automation
multimodal AI involving images, video, documents, and structured data
retrieval and knowledge systems
backend services supporting AI applications
evaluation and monitoring of production AI systems
internal AI tools for field and operations teams
We are looking for someone who enjoys building things, experimenting quickly, debugging real systems, and understanding how AI behaves outside controlled datasets.
You do not need to be a senior ML researcher or have experience with every technology listed below.
Strong fundamentals, curiosity, engineering ability, and evidence that you have actually built things matter more.
What You Might Work On
Computer Vision
Train, fine-tune, evaluate, and improve models for detecting road defects, infrastructure assets, objects, and other conditions from real-world imagery and video.
This may involve YOLO or similar detection models, segmentation, image classification, tracking, and multimodal vision models.
Video and Image Processing
Work with continuous footage collected from moving vehicles.
You may deal with:
frame extraction and sampling
image quality
motion blur
changing lighting
camera positioning
resolution and compression
object tracking across frames
duplicate detection
geo-referencing observations
The data will not always be clean. Part of the job is figuring out what can realistically be extracted from it.
Training Data and Evaluation
Help us improve datasets rather than treating model training as a black box.
This includes:
reviewing labels
defining annotation standards
identifying bad or ambiguous training examples
handling class imbalance
finding failure cases
creating train, validation, and test datasets
comparing models and thresholds
measuring precision and recall
checking whether a new model is actually better before deploying it
Agentic AI and LLM Systems
You may also work on AI agents and LLM-powered workflows around the platform.
Examples could include:
agents that analyse infrastructure findings
automated report generation
natural-language querying of survey data
document and knowledge retrieval
AI workflows that combine databases, APIs, images, maps, and documents
structured extraction from unstructured information
tool-calling agents
evaluation of LLM and agent behaviour
Experience with frameworks such as LangGraph, LangChain, CrewAI, OpenAI Agents SDK, Google ADK, or similar is useful, but understanding the underlying concepts matters more than knowing a particular framework.
Data Pipelines
Help process large amounts of imagery, video, metadata, and model outputs.
You may work on:
Python data pipelines
APIs
databases
cloud storage
batch processing
dataset versioning
GPU workloads
model inference pipelines
You should be comfortable investigating why a pipeline is slow, why data is missing, or why an output does not make sense.
Production AI
Models are useful only when the rest of the system works.
You may help with:
deploying models
building inference APIs
Docker
cloud GPUs
logging and monitoring
debugging production failures
improving inference speed and cost
connecting AI systems with existing applications
Core Requirements
Strong Python fundamentals.
Practical experience with machine learning, computer vision, LLMs, agentic AI, or a combination of these.
Ability to understand and modify existing code rather than only use no-code tools.
Familiarity with PyTorch, TensorFlow, or another ML framework.
Comfortable working with APIs, structured data, and basic backend systems.
Ability to debug problems independently.
Comfortable using Git.
Willingness to work hands-on with imperfect real-world datasets.
Ability to explain technical findings clearly to engineers and non-technical team members.
Comfortable communicating in English and spoken Hindi.
Willingness to occasionally visit field locations and understand how data is actually being collected.
Particularly Relevant Experience
Any of the following would strengthen your application:
YOLO or other object detection models.
Image segmentation.
Multi-object tracking.
OpenCV.
Vision Transformers or multimodal models.
PyTorch.
Fine-tuning models on custom datasets.
LLM APIs such as OpenAI, Anthropic, Gemini, or open-source models.
RAG systems.
AI agents and tool calling.
LangGraph, LangChain, CrewAI, Google ADK, OpenAI Agents SDK, or similar frameworks.
FastAPI, Flask, or Python backend development.
MongoDB, PostgreSQL, vector databases, or similar systems.
Docker and Linux.
AWS, Azure, or GCP.
GPU-based model training or inference.
Hugging Face.
ONNX or TensorRT.
MLflow, Weights & Biases, DVC, or similar tools.
You do not need experience with all of these.
Someone who is strong in computer vision but has never built an AI agent can still be a good fit.
Someone who has built strong LLM and agentic systems and has some ML or vision exposure can also be a good fit.
The Kind of Person We Are Looking For
We value builders.
You may be a good fit if you are the kind of person who has:
trained a model because you were curious whether it would work
built an AI agent that actually calls tools rather than just chatting
collected or labelled your own dataset
deployed a model or API yourself
built something for a college competition, hackathon, robotics team, or startup
used open-source models and experimented beyond tutorials
spent a night debugging something because you wanted to understand why it was failing
participated in SAE BAJA, Formula Student, robotics, drone, autonomous vehicle, computer vision, or similar engineering projects
A strong GitHub profile, personal project, competition project, internship, freelance build, or serious college project can matter as much as formal work experience.
Experience Level
We are open to candidates at different stages of their careers.
You could be:
an engineer with 1 to 4 years of experience
a strong recent graduate
someone coming from a startup or applied AI role
someone with a strong engineering or computer science background who has built unusually good projects
We care more about what you can build and how you think than the number of years written on your CV.
What You Will Get
Real production work. Your models and systems will work on actual field data and be used by operational teams.
Broad AI exposure. Work across computer vision, machine learning, multimodal AI, LLMs, agents, and production systems.
Direct access to the founders and senior engineers. You will be close to product and technical decisions rather than several layers away from them.
Fast learning. You will work on problems involving software, AI, hardware, field operations, data, and real customers.
Serious compute. Access to GPU infrastructure and the hardware required for ML experimentation.

Work arrangement
No