# Ankur Sharma, PhD

> Senior ML & Agentic AI Engineer building reasoning-traceable AI systems for genomics and biomedicine. PhD, 8+ years production experience. Based in Singapore.

## About

Ankur Sharma is a Senior ML & Agentic AI Engineer with a PhD in Biological Sciences from Nanyang Technological University (NTU), Singapore. He has 8+ years of experience shipping production machine learning systems, multi-agent AI, and cloud infrastructure across clinical diagnostics and genomics.

He specialises in building multi-agent AI systems — autonomous software where multiple AI agents collaborate to solve complex scientific and engineering problems. His work spans agentic AI for variant interpretation, production multi-agent systems for pipeline diagnostics, and self-optimising cloud genomics infrastructure.

He is open to Senior/Staff ML Engineer, Agentic AI Engineer, Research Engineer, and Applied Scientist roles. Based in Singapore (PR), open to remote or relocation.

## Contact

- Email: ankurs103@gmail.com
- LinkedIn: https://www.linkedin.com/in/ankurit/
- GitHub: https://github.com/ankurgenomics
- Portfolio: https://ankurgenomics.github.io/agentic-genomics/
- Resume (PDF): https://ankurgenomics.github.io/agentic-genomics/assets/Ankur_Sharma_Resume_2026.pdf

## Open-Source Projects

### agentic-genomics (GenomicsCopilot)

- URL: https://github.com/ankurgenomics/agentic-genomics
- License: MIT
- Language: Python 3.11+
- Stack: LangGraph, Claude/Anthropic API, Pydantic v2, pysam, Streamlit, Typer CLI

An open-source LangGraph agent for reasoning-traceable variant interpretation. Takes a VCF file and HPO phenotype terms and returns a ranked, explainable report of candidate genetic variants. The system has 7 deterministic nodes: VCF ingest, gnomAD/ClinVar/SpliceAI annotation via MyVariant.info, frequency filtering, Phrank-style HPO semantic-similarity scoring, ACMG-lite classification (PVS1 with LoF-intolerance gating, Richards et al. 2015 combining rules), LLM synthesis, and an LLM critic that fact-checks claims against the evidence JSON. Every run produces a machine-readable reasoning trace. This is a research demonstration, not a clinical tool. See LIMITATIONS.md for an honest accounting of scope.

### genomics-skills (Agent-Callable Skill Library)

- URL: https://github.com/ankurgenomics/genomics-skills
- License: MIT
- Language: Python 3.9+
- Stack: Python, Claude Haiku (LLM routing), cBioPortal/MyVariant/NCBI/PDB APIs, Pandas, Matplotlib

8 pure-Python agent-callable genomics skills: TCGA pan-cancer expression (9,479 real patient samples across 31 cancer types via cBioPortal), Kaplan-Meier survival analysis (Cox PH regression), GO/KEGG pathway enrichment, PubMed literature search, protein variant mapper, 3D protein structure viewer, and volcano plots. Each skill has a SKILL.md contract, CLI entrypoint, and deterministic output (TSV + PNG/SVG). LLM-powered routing via Claude Haiku maps natural-language queries to the right skill. Parquet caching for instant repeat queries.

## Production Work

### GenomicsOps AI (Personal Project)

A 5-agent system that autonomously triages DRAGEN, ICA, and SGE/HPC genomic pipeline failures. Agents: Trigger, Log Fetcher, RAG, Classifier, JIRA Writer. Reduced Mean Time To Resolution by 67% (3 days to 2 hours). Personal side project built on weekends. Built with multi-agent orchestration, Claude API, RAG, Python, JIRA/Confluence APIs.

### Autonomous Genomic Pipelines

Self-optimising WGS/RNA-seq workflows on AWS for a national precision medicine programme. Processed 6,000+ samples with minimal human intervention. Results: 40% lower compute costs, 50% less storage, 400 TB of genomic data managed. Stack: Nextflow (DSL2), AWS Batch, Lambda, Step Functions, Docker, Infrastructure as Code.

## Technical Skills

### Core
- Agentic AI: LangGraph, Claude/Anthropic API, Multi-Agent Orchestration, RAG, Tool-Calling Agents
- ML: Python, scikit-learn, PyTorch, Pandas/NumPy, Statistical Modelling, Feature Engineering
- Cloud: AWS (Batch, Lambda, S3, Step Functions), Nextflow, Docker, GitHub Actions, IaC
- Genomics: NGS (WGS/WES/RNA-seq/ChIP-seq), DRAGEN, GATK, Variant Interpretation, ACMG Guidelines

### Frequent
- OpenAI API, RAG & Vector Stores, MLflow, Kubernetes, R/Bioconductor, Snakemake
- Regulatory: DVT, Clinical Validation (LOD, Sensitivity, Specificity), FDA Compliance, cGMP, HIPAA

## Education

- PhD in Biological Sciences — Nanyang Technological University, Singapore (2016–2021). Thesis: Age-dependent transcriptional and epigenetic alterations in mouse hepatocytes. DOI: 10.32657/10356/155390
- M.Tech in Biotechnology — Anna University, India (2013–2015)
- B.Tech in Biotechnology — UPTU, India (2007–2011)

## Publications

- Sharma, A. (2021). Age-dependent transcriptional and epigenetic alterations in mouse hepatocytes. PhD thesis, NTU. DOI: 10.32657/10356/155390
- Kumari M, Taritla S, Sharma A, Jayabaskaran C. (2018). Antiproliferative and antioxidative bioactive compounds in extracts of marine-derived endophytic fungus. Frontiers in Microbiology.
- Sharma A. (2019). Significance of hepatocyte polyploidization in liver physiology and pathology. Conference poster, Cell Symposia, Chicago.

## Keywords

Ankur Sharma, agentic AI engineer, LangGraph developer, multi-agent systems, genomics AI, bioinformatics ML engineer, variant interpretation AI, clinical genomics, Singapore ML engineer, production ML, reasoning-traceable AI, open-source genomics tools, ACMG classification, LLM agent developer, Claude API, AWS genomics pipelines, Nextflow, research engineer Singapore