Data Scientist

Position: Data Scientist – LLM & Applied AI

Experience:

Employment Type:

Domain:

Reporting To:

Role Summary

highly hands-on Data Scientist

not purely academic

Key Responsibilities

1. LLM Research, Evaluation & Selection

• Evaluate, benchmark, and compare

  open-source LLMs

  (LLaMA-2/3, Mistral, Mixtral, Falcon, Qwen, Phi, etc.) and

  commercial LLMs

  (OpenAI, Anthropic, Google, Azure).
• Select appropriate models based on

  latency, accuracy, cost, explainability, and data-privacy requirements

  .
• Maintain an internal

  LLM capability matrix

  mapped to specific business use cases.

2. Prompt Engineering & Reasoning Design

• Design, test, and optimize

  prompt strategies

  :
• Zero-shot, few-shot, chain-of-thought (where applicable)
• Tool-calling and function-calling prompts
• Multi-agent and planner-executor patterns
• Build

  domain-aware prompts

  for healthcare workflows (clinical notes, scheduling, RCM, patient communication).
• Implement prompt versioning, prompt A/B testing, and regression checks.

3. Applied ML & Model Development

• Build and fine-tune

  ML/DL models

  (classification, NER, summarization, clustering, recommendation).
• Apply

  traditional ML + LLM hybrids

  where LLMs alone are not optimal.
• Perform feature engineering, model evaluation, and error analysis.
• Work with

  structured (SQL/FHIR)

  and

  unstructured (text, audio)

  data.

4. Local LLM & On-Prem Deployment

• Deploy and optimize

  local LLMs

  using frameworks such as:
• Ollama, vLLM, llama.cpp, HuggingFace Transformers
• Implement

  quantization (4-bit/8-bit)

  and performance tuning.
• Support

  air-gapped / HIPAA-compliant

  inference environments.
• Integrate local models with microservices and APIs.

5. RAG & Knowledge Systems

• Design and implement

  Retrieval-Augmented Generation (RAG)

  pipelines.
• Work with vector databases (FAISS, Chroma, Weaviate, Pinecone).
• Optimize chunking, embedding strategies, and relevance scoring.
• Ensure traceability and citation of retrieved sources.

6. AI System Integration & Productionization

• Collaborate with backend and frontend teams to integrate AI models into:
• Spring Boot / FastAPI services
• React-based applications
• Implement monitoring for

  accuracy drift, latency, hallucinations, and cost

  .
• Document AI behaviors clearly for BA, QA, and compliance teams.

7. Responsible AI & Compliance Awareness

• Apply

  PHI-safe design principles

  (prompt redaction, data minimization).
• Understand healthcare AI constraints (HIPAA, auditability, explainability).
• Support human-in-the-loop and fallback mechanisms.

Required Skills & Qualifications

Core Technical Skills

• Strong proficiency in

  Python

  (NumPy, Pandas, Scikit-learn).
• Solid understanding of

  ML fundamentals

  (supervised/unsupervised learning).
• Hands-on experience with

  LLMs (open-source + commercial)

  .
• Strong command of

  prompt engineering techniques

  .
• Experience deploying models locally or in controlled environments.

LLM & AI Tooling

• HuggingFace ecosystem
• OpenAI / Anthropic APIs
• Vector databases
• LangChain / LlamaIndex (or equivalent orchestration frameworks)

Data & Systems

• SQL and data modeling
• REST APIs
• Git, Docker (basic)
• Linux environments

Preferred / Good-to-Have Skills

• Experience in

  healthcare data

  (EHR, clinical text, FHIR concepts).
• Exposure to

  multimodal AI

  (speech-to-text, text-to-speech).
• Knowledge of

  model evaluation frameworks

  for LLMs.
• Familiarity with

  agentic AI architectures

  .
• Experience working in

  startup or fast-moving product teams

  .

Research & Mindset Expectations (Important)

• Strong inclination toward

  applied research

  , not just model usage.
• Ability to read and translate

  research papers into working prototypes

  .
• Curious, experimental, and iterative mindset.
• Clear understanding that

  accuracy, safety, and explainability

  matter more than flashy demos.

What We Offer

• Opportunity to work on

  real production AI systems

  used in US healthcare.
• Exposure to

  end-to-end AI lifecycle

  : research → prototype → production.
• Work with

  local LLMs, agentic systems, and multimodal AI

  .
• High ownership, visibility, and learning curve.

kvyas@omnimd.com

kvyas@omnimd.com