Machine Learning Engineer

Devsinc is hiring a skilled AI & ML Engineer with more than 2 years of professional experience in building and fine-tuning Generative AI models (LLMs, Diffusion Models), Vision-Language Models (VLMs), and both classical and deep learning systems, developing solutions from scratch and taking them end-to-end into production.
This role combines modeling and MLOps expertise , involving end-to-end ownership from model training and fine-tuning to optimization, deployment, and serving .
You’ll work on diverse, high-impact projects such as Generative AI applications, Stable Diffusion, OCR, theft detection, and recommendation systems , designing, optimizing, and serving custom models for real-world production use.
Key Responsibilities: Develop production inference stacks : Convert and optimize models (Torch → ONNX → TensorRT), quantize/prune, profile FLOPs and latency, and deliver low-latency GPU inference with minimal accuracy loss.
Build robust model-serving infrastructure : Implement FastAPI/gRPC inference services, token or frame-level streaming, model versioning and routing, autoscaling, rollbacks, and A/B testing.
Create Computer Vision solutions from scratch : Design pipelines for object detection, theft detection, OCR (document parsing, structured extraction), and surveillance analytics; fine-tune Hugging Face pretrained models when beneficial.
Fine-tune Stable Diffusion and other generative models for brand- or style-consistent image generation and downstream vision tasks.
Train and fine-tune Vision-Language Models (VLMs) for multimodal tasks (captioning, VQA, multimodal retrieval) using both from-scratch and transfer-learning approaches.
Design and adapt LLM-based Generative AI systems for conversational agents, summarization, RAG pipelines, and domain-specific fine-tuning.
Implement MLOps / LLMops / AIOps practices : Automate CI/CD for training and deployment, manage datasets and experiments, maintain model registries, and monitor latency, drift, and performance with alerting and retraining pipelines.
Develop data acquisition & ingestion pipelines : Build compliant scrapers, collectors, and scalable ingestion systems with proxy rotation and rate-limit handling.
Integrate third-party models and APIs (Hugging Face, OpenAI, etc.
) and design hybrid inference strategies combining local and cloud models for optimal performance.
Education : Bachelor’s or Master’s degree in Computer Science, Artificial Intelligence, or related field.
Experience : 2+ years of professional experience in AI/ML or relevant domains, with a proven track record of developing, training, and deploying machine learning or deep learning models in real-world environments.
Excellent understanding of classical ML (scikit-learn): regression, classification, clustering; able to design experiments and baselines.
Strong expertise in Computer Vision : object detection, segmentation, OCR pipelines (training from scratch and transfer learning).
Deep understanding of model optimization : quantization, pruning, distillation, FLOPs analysis, CUDA profiling, mixed precision, and inference performance trade-offs.
Proven ability to design and train models from scratch, (not only using pretrained checkpoints): architecture design, loss functions, training loops, and evaluation.
Hands-on experience with LLMs and diffusion-based models (e.
g., Stable Diffusion).
Proficiency with ONNX, TensorRT, TorchScript , and serving frameworks (Triton, TorchServe, or ONNX Runtime).
Skilled in GPU programming and CUDA optimization (profiling with nvprof/nsight, memory management, multi-GPU setups).
Strong backend engineering in Python (FastAPI, Flask), async programming, WebSockets/SSE, and RESTful API design.
Experience with containerization and orchestration (Docker, Kubernetes, Helm) and deploying GPU workloads to AWS/GCP/Azure or on-prem clusters.
Solid software engineering discipline : CI/CD, testing, code reviews, reproducibility, and version control.
Nice-to-Have: Familiarity with privacy-preserving ML (differential privacy, federated learning) and observability tools like Prometheus, Grafana, Sentry, or OpenTelemetry.
Collaborative – open to knowledge-sharing and teamwork.
Team Player – willing to support peers and contribute to collective success.
Growth Minded – eager to learn, improve, and adapt to emerging technologies.
Adaptable – flexible in dynamic, fast-paced environments.
Customer-Centric – focused on delivering solutions that create real business value.