Daxa, Inc
Website:
daxa.ai
Job details:
Senior ML Engineer
AI Governance & Security · DAXA · Pune, India or US (remote-friendly) · 5+ years
DAXA provides a data-first security and governance platform for AI applications and agents. This is an ambitious team shipping into large, demanding enterprises.
What you'll own
• Own the end-to-end ML systems that power the DAXA platform: data, training, evaluation, inference and deployment.
• Build and evolve training and fine-tuning pipelines for small language models and encoder-family classifiers.
• Design evaluation systems that measure capability, robustness, safety and real-world product performance.
• Architect high-performance inference systems, optimizing latency, GPU utilization, memory, cost and reliability.
• Build data pipelines and systems for high-quality real-world and synthetic training data.
• Establish reliable production infrastructure for deploying, monitoring and continuously improving models, including on-prem and air-gapped customer environments.
• Advance our detection surface: sensitive-data and entity detection, prompt injection, data exfiltration, and anomalous agent behavior.
• Partner closely with product and application engineering to turn model capabilities into product improvements.
• Make pragmatic technical trade-offs and iterate rapidly based on real-world performance.
What we're looking for
Must have
• 5+ years hands-on in ML/NLP, with production models you personally shipped and maintained.
• Real fine-tuning depth on transformer encoders for both sequence and token classification.
• Strong Python; fluent in PyTorch and Hugging Face Transformers.
• Dataset ownership: labeling schemes, curation, and synthetic data generation for long-tail classes.
• Rigorous evaluation discipline, including managing precision/recall trade-offs in production.
• Model compression and serving optimization (distillation, quantization, ONNX, Triton, TensorRT), including CPU-only inference where no GPU is available.
• Taking a model from notebook to a latency-constrained production service under a hard p95 budget.
Strong plus
• AI security or adversarial ML: prompt injection, red-teaming, evasion.
• Data governance, DLP, IAM, or privacy/compliance context (GDPR, HIPAA, PCI).
• RAG, embeddings, retrieval, or knowledge graphs.
• Multilingual NLP, or shipping and versioning models into regulated, sovereign or air-gapped environments with no egress.
How we work
The technical bar is table stakes. These are the things that actually determine whether someone does well here.
• High agency. You'll get a customer symptom rather than a spec. Find the real problem, decide, ship.
• AI-native by default. Coding agents and frontier models are a core part of how you work. We care about what you ship and how well you understand it.
• Write things down. Design docs, eval writeups, decision records. We're distributed across time zones.
• Talk to customers. You'll sit on calls with enterprise security teams and CISOs.
• Own it to production. You build it, you ship it, you watch it in production, you fix it.
• Comfort with ambiguity. Startup pace, shifting priorities, incomplete information. Strong opinions, loosely held, low ego.
Why this role
You'd be the senior ML voice in the company, setting direction on the classifier architecture rather than executing someone else's roadmap. The problem space is genuinely unsolved: governing agent behavior is being defined right now, in public, and what you build here lands in production at named enterprise accounts within weeks, not quarters.
- If you want to build the intelligence that makes enterprise AI trustworthy, and you want your work in front of real users fast, we'd like to talk.
Click on Apply to know more.
