Deep Learning Engineer

• *Job Title:** Deep Learning Engineer (Mid - Senior Level) • *Location:** Kolkata • *Experience:** 2 - 5+ years • *Salary:** No bar for the right candidate • *Employment Type:** Full-time • *Role Overview:**

We are looking for a Deep Learning Engineer who can move fluidly between theory and production. This role requires someone who understands not just how to use models, but how they work, when they fail, and how to fix or redesign them. You will work on problems spanning computer vision and natural language processing, and will be expected to design, train, debug, and deploy models that solve real-world tasks under practical constraints. • *Key Responsibilities:**

\* Design, implement, and optimize deep learning models for CV and NLP tasks

\* Translate ambiguous business/product requirements into well-defined ML problems

\* Build custom architectures when standard approaches are insufficient

\* Debug training instability, convergence issues, and performance bottlenecks

\* Develop and experiment with custom:

\* Loss functions

\* Activation functions

\* Training strategies

\* Work with large-scale datasets: cleaning, preprocessing, augmentation, and validation

\* Evaluate models rigorously using appropriate metrics and error analysis

\* Deploy models as production-ready APIs (e.g., using FastAPI or similar frameworks)

\* Monitor and iterate on models post-deployment • *Required Skills & Qualifications:**

Deep Learning Fundamentals • *Strong understanding of:** • CNNs (architectural variants, feature extraction, transfer learning) • RNNs, LSTMs (sequence modeling, limitations, alternatives) • *Ability to reason about:** • Gradient flow, vanishing/exploding gradients • Optimization dynamics • Regularization techniques • Experience debugging model behavior beyond tuning hyper parameters • *Domains:**

Solid hands-on experience in: • Computer Vision (classification, detection, segmentation, etc.) • NLP (sequence modeling, embeddings, transformers basics) • *Programming & Engineering:** • Strong Python programming skills • Experience with PyTorch / TensorFlow (preferably PyTorch-heavy) • Ability to write clean, maintainable, and testable code • Experience building APIs for model serving (e.g., FastAPI) • *Experimentation & Tooling:** • Familiarity with experiment tracking and visualization tools: • ML flow • Tensor Board or equivalent tools • Strong experimentation discipline (reproducibility, logging, ablations) • Data Understanding • **Strong intuition for data:** • Distribution shifts • Noise handling • Feature relevance • Ability to identify when the problem is data, not model • Modern ML Systems • *Working knowledge of:** • Large Language Models (LLMs) • Vision-Language Models (VLMs) • Familiarity with fine-tuning, prompt-based methods, or adapters is a plus • *Nice to Have (Bonus Skills):**

CUDA programming or GPU kernel-level optimization • *Experience with model optimization:** • Quantization, pruning, distillation • Familiarity with deployment stacks (Docker, Kubernetes) • Experience with vector databases or retrieval systems • Knowledge of distributed training

If you're the type who reads papers, re-implements ideas, breaks things intentionally to understand them, and then ships robust systems - this role will suit you.