Senior AI/ML Engineer
Kathmandu, Nepal2w ago
Key details
- Location
- Kathmandu, Nepal
- Work style
- Onsite
- Level
- Senior
- Job type
- Full-time
- Experience
- 3–6 yrs
- Function
- Data & ML
- Focus area
- AI / ML
- Posted
- 2w ago
Skills
About the role
About Role
As a Senior AI/ML Engineer at Leapfrog Technology, you will build Generative AI applications end to end, on a strong base of core machine learning. You will decide when AI is the right tool, take AI features from problem framing to production and beyond, and lead a small team while doing it.
Problem Framing and AI Judgment: Define the problem and the success metric with product and clients. Decide whether AI is needed at all by comparing rules, SQL, classical ML, and LLMs, and document why.
End-to-End Gen AI Delivery: Integrate LLM APIs. Build the APIs and the frontend with AI coding tools. Design the AI user experience: streaming responses, citations, and capturing user feedback. Use structured outputs and tool calling, and choose models on quality, cost, and latency.
RAG, Workflows, and Agents: Design RAG pipelines end to end: ingestion, chunking, embeddings, vector store, retrieval, and reranking. Build workflows and agents with Strands Agents or LangGraph, MCP, and Amazon Bedrock AgentCore, and pick the simpler one when it is enough.
Guardrails and Security: Validate and filter outputs, and add fallback, retry, and timeout logic. Protect against prompt injection and data leakage, especially for healthcare data.
Evaluation: Define "good" before you build. Create evaluation sets and metrics for every model, RAG pipeline, workflow, and agent. Version prompts as code, calibrate LLM-as-judge against human labels, and gate releases with evaluation in CI/CD.
Deployment and Observability: Ship through CI/CD. Trace and monitor latency, cost per query, token usage, accuracy, and drift.
Continuous Improvement: Turn production failures into new evaluation cases, and fix the root cause.
Technology Evaluation: Track new models and tools, and test them against our evaluation sets before adopting them.
Classical ML: Reason when classical ML beats an LLM (cost, latency, tabular data). Then formulate the problem, define the metric, and train, evaluate, and deploy the model.
Team Leadership: Lead a small team on a feature or workstream: plan, estimate, unblock, track risks, and deliver. Review code and designs, and mentor junior engineers.
Cross-functional Collaboration: Work with clients, product managers, and engineers, and explain AI trade-offs and limits in plain language.
Organizational Impact: Present in knowledge-sharing sessions (KSS), and turn project work into reusable assets with the AI Center of Excellence (CoE).
Experience and Education
3+ years in software or ML engineering, with LLM features shipped to production (not only prototypes).
A degree in Computer Science, Engineering, AI, Data Science, Applied Mathematics, or a related field. A Master's degree is a plus.
Technical Skills
Core ML: Supervised and unsupervised learning, ensembles such as gradient boosting, and neural networks, including transformers. Working knowledge of scikit-learn and PyTorch. Depth in one of NLP or Computer Vision is preferred.
ML problem solving: Ability to judge when classical ML fits better than an LLM, formulate the problem (target, features, data, metric), and run the full cycle (train, evaluate, deploy, monitor), while spotting overfitting, data leakage, and class imbalance.
LLM integration: Hands-on experience with LLM APIs, prompt and context design (system prompts, few-shot examples, structured outputs), and a clear view of when fine-tuning is worth its cost. Experience with multimodal inputs, such as images and documents.
RAG: Ingestion, chunking, embeddings, vector stores or Amazon Bedrock Knowledge Bases, retrieval, and reranking.
Workflows and agents: Tool calling with Strands Agents or LangGraph, MCP, and Amazon Bedrock AgentCore, context and memory management for long or multi-turn tasks, and the judgment to avoid an agent when a workflow is enough.
Guardrails: Output validation, content filtering (for example Amazon Bedrock Guardrails), fallback and retry, defenses against prompt injection and data leakage, least-privilege tool access, and human approval for high-risk agent actions.
Evaluation: Ability to design evaluation for any ML or Gen AI system: build representative test sets, choose metrics that reflect real success, and compare against a baseline. Evaluate both individual components and end-to-end behavior, and use automated, model-based, and human evaluation where each fits. Gate releases with regression tests, monitor quality in production, and confirm that improvements are real before trusting them.
Observability: Tracing across LLM calls, tool calls, and agent steps, plus dashboards and alerts for latency, cost, tokens, errors, and quality (Langfuse, OpenTelemetry, or Amazon CloudWatch).
Software engineering: Strong, production-quality Python, and Python services with FastAPI (API design, authentication, rate limiting, async LLM calls). Cost control for LLM services: caching, model routing, and token budgets. Git, Docker, CI/CD, and SQL or NoSQL databases.
AI-assisted engineering: Full features across the stack, including frontend, built with AI coding tools (Claude Code, Cursor, or GitHub Copilot), kept safe with small changes, tests, and review.
Cloud: AI workloads on AWS (Amazon Bedrock, Lambda, ECS, S3).
Domain: Awareness of healthcare data privacy (HIPAA, PHI), or willingness to learn quickly. Most of our client work is in digital health.
Soft Skills
Ability to lead a small team on a feature, and to mentor engineers and give useful code review.
Excellent written and verbal communication in English, and comfort in client conversations.
Ability to explain uncertainty and AI limits honestly to non-technical people.
Sound technical decisions, backed by evidence.
Ability to find risks early and do root-cause analysis when things go wrong.
Experience with Agile methodologies like Scrum or Kanban.
Good documentation skills.
🌞 Life at Leapfrog:
Work with talented engineers and remarkable professionals in Nepal, the US, and Europe, building state-of-the-art digital health products.
Have an amazing career growth plan with support from experienced mentors and unlimited learning opportunities.
Be a part of an experienced team—apply the best industry practices, get mentored by experienced professionals, and be ready to mentor the new ones.
Get exposure to clients who are leading entrepreneurs and startups from Silicon Valley.
Your work will be at the cutting edge of healthcare innovation and make a difference in the lives of thousands of people worldwide.
Get growth opportunities, not just in your work. We organize various internal initiatives where you can be a part of a bigger learning community with webinars, knowledge-sharing sessions (KSS), and internal conferences.
Be a part of a safe workplace where it’s okay to fail fast— get constructive feedback and grow as you move ahead.
Receive recognition for your contributions and the positive impact you create. We always look for opportunities to celebrate our efforts, play games, and have fun together!
You will have a lot of fun at Leapfrog— there’ll never be a dull day.
🎁 Benefits and perks:
We believe our people are our greatest strength. With perks and benefits, we intend to offer everything our people need to do their best while maintaining a healthy balance between work and personal life.
Daily Work Support
💻 Support for resources: Receive a monthly internet and laptop allowance, and request resources for convenient WFH.
📆 5-day workweek: Enjoy weekends off and return refreshed on Mondays.
🍔 Food: Access to healthy meals while on-site.
Professional Growth
📚 'Growth mindset' fund: Investment in your post-grad degree, learning opportunities, and career growth.
👶 Child scholarship: Scholarships for your child in selected schools.
📈 Bi-annual bonus: Get timely festival and company bonuses.
💸 Advance salary: Request salary in advance.
Work/Life Balance
✈️ Paid time-offs: Annual, sick, compensation, and bereavement leaves.
🍼 Maternity leave: 3 months of paid leave, 1 month of unpaid leave, and 90 days of reduced workload.
🐣 Paternity leave: 11 days of paid paternity leave and reduced workload after.
💊 Menstruation leave: 6 days of paid menstruation leave annually.
We take care of you
🌿 Your well-being: Company-funded counseling services from certified therapists as well as COVID care services.
💰 Retirement fund: Social security fund with health and accidental insurance.
Celebrating your efforts
🎉 Engagement and Events: Team lunch/dinner, company outings, and festival celebrations with your favorite co-workers!
😎 Exposure: Visit and work in the States and build better client rapport.
💖 Recognition: In-house awards, appreciation platforms, and 5th-anniversary gifts!
How well do you match this job?
Free account needed to apply or save. It takes a minute, and we'll bring you right back to this job.