Akilan Chithra Sathish
Dallas, TXOpen to SWE/ML internships

MLOps Engineer.

From prototyping models to orchestrating data, services, and infra.

End-to-end

Model to cloud, observability included

Real-time

Streams, ROS2 & fast backends

Ops-ready

Docker, CI/CD, infra as code

GitHubLinkedInEmailResume
10+
Projects
2
Research
Pytorch/TensorflowAWSReact/Next.jsTypeScriptPythonFastAPI/FlaskPyTorchROS2PostgreSQLDocker
Scroll to explore
Illusion background
Hello, world🌎

Experience

Recent work & research

Undergraduate Researcher • SharedAirDFW Team
UTD MINTS • Dr. David Lary • Dallas, TX
Jun 2024 – Present
  • Redesigned SharedAirDFW with Earth Engine + Geemap to give UTD ownership and simplify workflows.
  • Merged four microservices into one Flask + Leaflet + MQTT pipeline, cutting maintenance overhead.
  • Automated Quarto/Jupyter charting to double data-viz throughput for pollution research.
Software Developer • Vehicle Interface Team
UTD NOVA (Autonomous Driving) • Dallas, TX
Jan 2025 – Present
  • Designed ROS2 fusion module for GNSS/IMU/LiDAR to improve localization accuracy.
  • Reduced pose-estimation complexity ~50% with streamlined fusion algorithms.
  • Migrated GPS odometry to a hybrid pipeline, improving accuracy from ~3 m to < 1 m on 10-min drives.
Undergraduate Researcher • Maestro (Synthetic Audio for CLAP)
UTD ACM • Dallas, TX
Aug 2024 – Dec 2024
  • Showed synthetic audio can improve CLAP accuracy by ~6% in low-data settings.
  • Built PyTorch eval suite (mAP on OpenMIC-2018; cosine sim on GTZAN).
  • Created LLM prompt pipeline from a keyword corpus to boost dataset diversity.
Backend Dev Shadow • Full-Stack Team
CareerBoosts • Frisco, TX
Jun 2023 – Aug 2023
  • FastAPI + PostgreSQL ingestion handling 20k+ daily requests with low latency.
  • Flask + React dashboard in Docker improved reporting speed by ~40%.
  • GraphQL endpoint reduced over-fetching between backend and frontend.

About

Hey! I’m Akilan. I am a Junior at the University of Texas at Dallas studying Computer Science. I love building where autonomy, audio, and ML meet—and then pushing those ideas to cloud scale so they actually ship.

On the autonomy side I’ve built ROS2 sensor-fusion pieces (GNSS/IMU/LiDAR) and tooling that turns experiments into stable services. On the backend side I’ve consolidated microservices into cleaner pipelines (Flask + Leaflet + MQTT) and stood up FastAPI + Postgres stacks with solid DX and observability. In audio, I’ve shipped a web stem-separator (Demucs + Celery/Redis) and a melody-prediction RNN with response-time budgets in mind, plus a CLAP study exploring data-scarce training.

Research: I presented a poster on using synthetic data to train a CLAP model—and showed that, in low-data settings, synthetic examples can actually help the model learn useful audio-text alignments.

Right now: sketching agentic workflows for customer support—triage, suggested replies, and routing—so teams can do more with fewer human agents while keeping quality and feedback loops tight.

How I work: prototype fast, productionize intentionally, ship behind flags, add traces/metrics from day one, and write docs that future-me actually thanks me for.

Outside of work it’s music, soccer, and TV shows (Ozark is peak). I also watch football—nobody’s matching the Detroit Lions’ work ethic and team culture. Music-wise I bounce between old ’80s and modern picks like Lola Young and Twenty One Pilots—I switch genres like clothes.

When I need a reset, I speed-solve Rubik’s cubes and then get back to building.

Projects

A few focused builds

DJSplitter

DJSplitter

Featured

Web-based stem separation + two-deck mixer. Demucs backend with Flask/Celery/Redis; React/Vite UI with Tone.js.

TypeScriptReact/ViteTone.jsFlaskDemucsCeleryRedis
Details
MelodyGen

MelodyGen

Featured

RNN melody generator (TensorFlow) with a simple web demo; composes short sequences from prior notes.

TensorFlowPythonRNNJupyterFlask/Next.js (demo)
Details
TicTacToeAI

TicTacToeAI

Featured

Unbeatable Tic-Tac-Toe using MiniMax with alpha-beta pruning; clean reference implementation.

PythonMiniMaxAlpha-Beta
Details
GestureController

GestureController

Featured

Touch-free volume & brightness using webcam hand gestures. MediaPipe + OpenCV with a lightweight Flask bridge.

PythonMediaPipeOpenCVFlask
Details

Skills

Scroll to view faster

TypeScriptReactNext.jsPythonFastAPINode.jsPostgresMongoDBDockerKubernetesTypeScriptReactNext.jsPythonFastAPINode.jsPostgresMongoDBDockerKubernetesTypeScriptReactNext.jsPythonFastAPINode.jsPostgresMongoDBDockerKubernetesTypeScriptReactNext.jsPythonFastAPINode.jsPostgresMongoDBDockerKubernetesTypeScriptReactNext.jsPythonFastAPINode.jsPostgresMongoDBDockerKubernetesTypeScriptReactNext.jsPythonFastAPINode.jsPostgresMongoDBDockerKubernetes
TerraformGitHub ActionsAWSLangChainLangGraphMCPSageMakerGrafanaPrometheusFlinkTerraformGitHub ActionsAWSLangChainLangGraphMCPSageMakerGrafanaPrometheusFlinkTerraformGitHub ActionsAWSLangChainLangGraphMCPSageMakerGrafanaPrometheusFlinkTerraformGitHub ActionsAWSLangChainLangGraphMCPSageMakerGrafanaPrometheusFlinkTerraformGitHub ActionsAWSLangChainLangGraphMCPSageMakerGrafanaPrometheusFlinkTerraformGitHub ActionsAWSLangChainLangGraphMCPSageMakerGrafanaPrometheusFlink

LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•
LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•
LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•
LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•LET’S TALK•
Email me

© 2025 Akilan Chithra Sathish. Built with React.

GitHubLinkedInEmailResume