☕️About Me
Aakash Kaushik is a Software Engineer/Machine Learning Engineer at Lyric, focusing on machine learning, MLOps, and software engineering. With expertise in Generative AI, LLMs, VLMs, and cloud infrastructure, he builds scalable solutions that process millions of requests daily.
His technical skills span Golang, Python, and various cloud technologies (GCP, AWS, Azure), with a focus on high-performance distributed systems and MLOps. Aakash has contributed to open-source projects, including mlpack, and has experience building document processing pipelines, proxy servers for LLM providers, and cloud-agnostic infrastructure.
The name “Kausky” comes from Aakash (meaning “sky” in Hindi) and Kaushik (surname, “kau”) -> kausky. This represents his vision of building AI solutions with limitless possibilities.
Aakash holds a BTech in Computer Science Engineering from SRM University with a CGPA of 9.55/10 and has worked across multiple roles in AI/ML development.
- Machine Learning & Generative AI
- LLMs & VLMs
- MLOps & Cloud Infrastructure
- Distributed Systems
- API Gateway Development
- Performance Optimization
BTech in Computer Science Engineering
SRM University (Sri Ramaswamy Memorial University)
I’m an ML Engineer and Software Engineer 3 at Tune AI, specializing in developing and deploying machine learning models and distributed systems for real-world applications. My work focuses on creating robust AI solutions that solve complex problems and drive business value.
Key Areas of Expertise:
- Generative AI & LLMs: Building high-throughput proxy servers handling over 1M requests/day for OpenAI, Anthropic, and other LLM providers
- Document Processing: Developing multimodal extraction pipelines processing 100K+ documents daily with ~95% precision/recall
- Cloud Infrastructure: Creating flexible, scalable infrastructure across AWS, GCP, and Azure using Kubernetes and IaC
- MLOps: Implementing robust monitoring, logging, and deployment pipelines for ML systems
The name “Kausky” comes from Aakash (meaning “sky” in Hindi) and Kaushik (surname, “kau”) -> kausky. This represents my vision of building AI solutions with limitless possibilities.
Feel free to reach out for collaboration opportunities! 😃
Languages
- Golang (2+ years): Microservices, API gateways, high-performance systems
- Python (4+ years): ML/AI development, data processing, API development
- Others: C++, Bash, SQL
Technologies
- Cloud: GCP, AWS, Azure, Kubernetes, Docker
- ML/AI: PyTorch, TensorFlow, OpenVINO, LLMs, VLMs
- Infrastructure: Pulumi, Terraform, CI/CD, gRPC, REST APIs
- Data: Redis, MySQL, PostgreSQL, Cloud Storage (S3, GCS, Azure Blob)
- mlpack 4: A fast, header-only C++ machine learning library
Published in the Journal of Open Source Software (2023)
DOI: 10.21105/joss.05026
Google Summer of Code - Mlpack
As a developer for Google Summer of Code with Mlpack, I:
- Implemented MobileNetV1 and ResNet model builders in C++, integrating pre-trained weights to reduce training time by 40%
- Added Mean Absolute Percentage Error (MAPE) and Softmin Activation function with backward implementation
- Spearheaded the migration of approximately 60% of core testing suite from Boost to Catch2
- Addressed over 100 static code analysis warnings and style issues