Applied AI engineer · Bengaluru, India
I design and ship focused AI products across spatial intelligence, deep learning, agentic systems, and cloud-native applications.
Second page / system map
Each block is a project, capability, or research direction in the autonomy stack.
Selected systems
Research experiments, visual systems, and autonomous software from the lab.
Spatial intelligence work focused on analysing the Western Ghats with NASA data and Sentinel-2 satellite imagery. The project explores large-scale environmental data, remote sensing, spatial representations, and computer-vision workflows for canopy and land-cover understanding.
A prototype for representing and detecting potential threats in a simulated 3D environment. It models spatial data, point clouds, voxelized structure, and 3D bounding-box relationships around an autonomous system.
A Gazebo-based digital-twin direction that updates a simulated environment using spatial information generated by a robotic system. It connects SLAM data, 3D spatial state, and simulation into a representation of the physical world.
A collection of focused experiments for learning spatial intelligence from first principles: LiDAR, SLAM, visual odometry, sensor fusion, neural perception, spatial modelling, and simulation. Each block isolates one capability used in autonomous systems.
An embedding-driven semantic graph engine that ingests text or web content, builds similarity edges, performs weighted walks, and generates cross-domain hypotheses with graph and walk visualizations.
An experimental pathfinding system that evolves from basic LLM generation to reward-based learning, LLM-as-judge evaluation, and Group Relative Policy Optimization. Candidate paths are generated, scored, compared, and refined in a grid-world environment.
A simulation-first autonomy stack for drone navigation. The system connects Webots environments to an action loop with Gemini Flash, WebSocket communication, speech-to-instruction input, custom controllers, YOLO-based perception, and computer-vision utilities.
A Java-based disease-detection application that packages a trained image-classification workflow into a mobile-oriented user experience, connecting model inference with a practical diagnostic interface.
Reinforcement-learning experiments for selecting effective image-preprocessing hyperparameters before classification, treating preprocessing as a search and optimization problem.
A hybrid movie-recommendation system combining content and collaborative signals to rank relevant titles for a user.
Engineering approach
Every model or agent earns its place by improving a real workflow, not by adding novelty.
Tracing, evaluation, visual outputs, and clear failure modes are part of the product from the first prototype.
Digital twins and small environments make autonomy, perception, and control loops testable before deployment.
About the builder
I'm Rakshith Srinivasan, an Associate Software Engineer working at the intersection of machine learning, developer tooling, spatial computing, and deployable software.
Education
Bachelor of Engineering in Information Science and Engineering at Dayananda Sagar Academy of Technology and Management.
Start a conversation
Open to applied AI, intelligent infrastructure, robotics, and early-stage deep-tech work.
Technical depth
The implementation details behind the project labels.
Sentinel-2 and GEDI/LiDAR data preparation, rasterization and tiling, embedding generation, similarity graphs, weighted random walks, semantic mutation, and visualization/report generation.
Explorations across SLAM, deep SLAM, visual odometry, LiDAR, voxel grids, tensor projection, sensor fusion, 3D modeling, PnP, AirSim, neural simulation, flight-training environments, and threat-scene topology.
Webots simulation with controller scripts, action planning, moving-platform control, Gemini/OpenAI-compatible LLM calls, Redis operations, WebSocket transport, Vosk speech recognition, VAD, PyAudio, and test modules for the voice and drone loop.
Production-facing work across AI-assisted product intelligence, RAG-oriented workflows, multi-agent orchestration, Azure infrastructure automation, FastAPI services, frontend integration, monitoring, and deployment tooling.