Experience

Hover a dot for dates, click to jump to the role.

// The chart lives in this cell, not with the definitions below: a node
// already displayed by another (hidden) cell is shown as an object, not drawn
{
  const vertical = width < 640;
  const tip = { fill: palette.panel, stroke: palette.navy };
  const channels = { Role: "title", When: "range" };
  // Page-coloured outline so a neighbouring stem passes behind a label, not through it
  const halo = { stroke: "#1b2432", strokeWidth: 5, paintOrder: "stroke" };
  // Desktop layout: stack labels into lanes by their measured pixel extent,
  // so nothing overlaps at any width. Labels that would run off the right
  // edge flip to the left of their stem.
  const margin = 20, pad = 10;
  const x = d3.scaleUtc([roles[0].start, d3.max(roles, (d) => d.end)], [margin, width - margin]);
  const ends = [];  // rightmost pixel used in each lane
  const laned = roles.map((d) => {
    const px = x(d.start);
    const w = pad + d3.max(d.label.split("\n"), measure);
    const left = px + w > width - margin;
    const [x0, x1] = left ? [px - w, px + 6] : [px - 6, px + w];
    const fits = (k) => ends[k] + pad < x0;
    // Prefer a lane whose stem also clears the labels below it
    let lane = ends.findIndex((_, k) => fits(k) && ends.slice(0, k).every((e) => e < px - 4));
    if (lane < 0) lane = ends.findIndex((_, k) => fits(k));
    if (lane < 0) lane = ends.length;
    ends[lane] = x1;
    return { ...d, level: lane + 1, left };
  });
  const lanes = ends.length;

  // Phone layout: years run down the page. Labels are laddered at least
  // `gap` apart and a bent stem ties each back to its true start date.
  const gap = 1.25 * 365.25 * 864e5;
  const placed = [];
  for (const d of roles) {
    const prev = placed.at(-1);
    placed.push({ ...d, at: new Date(Math.max(+d.start, prev ? +prev.at + gap : 0)) });
  }

  const chart = vertical
    ? Plot.plot({
        width, height: 960, marginTop: 16, marginLeft: 44, marginRight: 8,
        style: { background: "transparent", color: palette.sand, fontSize: "12px" },
        x: { axis: null, domain: [0, 6] },
        y: { type: "utc", reverse: true, grid: true, label: null,
             domain: [placed[0].start, new Date(+placed.at(-1).at + gap / 2)] },
        marks: [
          Plot.ruleX([0], { stroke: palette.navy }),
          Plot.ruleX(placed, { x: 0, y1: "start", y2: "end", stroke: color, strokeWidth: 4, strokeOpacity: 0.5 }),
          Plot.link(placed, { x1: 0, y1: "start", x2: 1, y2: "at", curve: "bump-x", stroke: color }),
          Plot.dot(placed, { y: "at", x: 1, r: 6, fill: color, stroke: palette.panel,
                             channels, tip: { ...tip, format: { x: false, y: false, fill: false } } }),
          Plot.text(placed, { y: "at", x: 1, text: "label", dx: 10, textAnchor: "start",
                              fill: "#f3ede2", lineHeight: 1.2, ...halo })
        ]
      })
    : Plot.plot({
        width, height: 100 + 70 * lanes, marginTop: 10, marginBottom: 30,
        marginLeft: margin, marginRight: margin,
        style: { background: "transparent", color: palette.sand, fontSize: "12px" },
        y: { axis: null, domain: [0, lanes + 0.6] },
        x: { type: "utc", label: null, grid: true, domain: x.domain() },
        marks: [
          Plot.ruleY([0], { stroke: palette.navy }),
          Plot.ruleY(laned, { y: 0, x1: "start", x2: "end", stroke: color, strokeWidth: 6, strokeOpacity: 0.5 }),
          Plot.ruleX(laned, { x: "start", y1: 0, y2: "level", stroke: color }),
          Plot.dot(laned, { x: "start", y: "level", r: 6, fill: color, stroke: palette.panel,
                            channels, tip: { ...tip, format: { x: false, y: false, fill: false } } }),
          Plot.text(laned, { filter: (d) => !d.left, x: "start", y: "level", text: "label", lineHeight: 1.2,
                             fill: "#f3ede2", dx: 10, textAnchor: "start", ...halo }),
          Plot.text(laned, { filter: (d) => d.left, x: "start", y: "level", text: "label", lineHeight: 1.2,
                             fill: "#f3ede2", dx: -10, textAnchor: "end", ...halo })
        ]
      });

  // Plot binds each circle's datum to its row index
  d3.select(chart).selectAll("g[aria-label=dot] circle")
    .style("cursor", "pointer")
    .on("click", (event, i) => jump(roles[i]));
  return chart;
}

Plot = import("https://cdn.jsdelivr.net/npm/@observablehq/plot@0.6.16/+esm")

roles = {
  const parse = d3.utcParse("%Y-%m");
  const now = new Date();
  const rows = await FileAttachment("data/experience.csv").csv();
  return rows.map((d) => ({
    ...d,
    start: parse(d.start),
    end: d.end ? parse(d.end) : now,
    label: `${d.company}\n${d.title}`,
    range: `${d.start.replace("-", "/")} – ${d.end ? d.end.replace("-", "/") : "present"}`
  }));
}

// Label width in pixels, in the page's own font (falls back to an estimate)
measure = {
  const ctx = document.createElement("canvas").getContext?.("2d");
  if (!ctx) return (s) => s.length * 7;
  ctx.font = `12px ${getComputedStyle(document.body).fontFamily}`;
  return (s) => ctx.measureText(s).width;
}

palette = ({ sage: "#a7b38b", clay: "#c99b80", sand: "#cfc6b4", navy: "#33425a", panel: "#243041" })

color = (d) => d.category === "current" ? palette.clay : d.category === "academia" ? palette.sand : palette.sage

// Scroll to the matching CV entry; Takeda appears twice, so match the title too
jump = (d) => {
  const hit = [...document.querySelectorAll("main section.level2, main section.level3")]
    .find((s) => s.querySelector("h2, h3")?.textContent.includes(d.anchor) && s.textContent.includes(d.title));
  hit?.scrollIntoView({ behavior: "smooth", block: "start" });
}

Experience

Korro Bio (Cambridge, MA)

Director of Data Science
August 2025 – Present

• Lead Korro’s Data Science function, a blended team of up to 3 on-site FTEs and 3 international contractors, owning hiring, performance management, budget, vendor contracts, and the cross-Discovery/Clinical roadmap.
• Established a cross-time-zone operating model (asynchronous review, documentation standards, shared code and data conventions) so a distributed team delivers reproducibly.
• Interrogated EHR-scale real-world datasets (NIH All of Us Researcher Workbench, NashBio) for indication expansion, eligibility and enrollment feasibility, and natural-history characterization; findings informed clinical trial design and were presented at EASL 2026 (see Fischer, Roy, Tapper 2026).
• Directed agentic-AI development of an integrated software ecosystem that unifies ADAR-recruiting oligonucleotide workflows from target identification and validation through ML/DL-guided design and automated high-throughput screening ingest and analysis.
• Materially reduced function spend (contractor, cloud, and SaaS) while expanding computational biology output, through agentic coding tools and a secure on-prem LLM deployment that keeps proprietary sequence and clinical data in-network.
• Championed adoption of AI coding agents across the organization, raising codebase quality and development velocity.

---

Affini-T Therapeutics (Watertown, MA; company closed 2025)

Associate Director: Precision and Translational Medicine and Computational Biology
December 2021 – July 2025

• Led multidisciplinary initiatives at a cell and gene therapy start-up, successfully driving translational and computational strategies supporting AFNT-211, a multi-site single-arm TCR-T clinical trial in solid tumors.
• Led a diverse Computational team of 1 FTE, 3 contractors, 2 associated Discovery scientists, 2 academic consultants, and multiple vendors.
• Integrated and harmonized complex multi-source clinical, translational, and manufacturing data types, including EDC, CMC, ctDNA, PK/PD, flow cytometry, scRNA-seq, and cytokines.
• Designed and maintained interactive dashboards that increased trial data accessibility, stakeholder engagement, and decision making.
• Developed statistical models to predict responses based on enrollment criteria and improve patient trial response rates.
• Conducted site visits engaging clinical investigators, leading to increased in enrollment rates and enhanced site feedback incorporation.
• Contributed patient population analysis based on preclinical experiments for inclusion in IND filings.
• Created a novel patient population estimation algorithm based on HLA and mutation frequencies (see Roy2025 for HLA component).
• Leveraged public and proprietary datasets to assess HLA Loss of Heterozygosity impacts on therapy efficacy (see ASGCT 2024 abstract and Poster).
• Architected and maintained a multi-account AWS cloud infrastructure supporting both Discovery and Clinical programs.
• Championed deployment of computational workflow management systems enabling scalable, reproducible analysis pipelines.
• Supported Discovery and Prelinical activities from 2021–2024.
• Led efforts in TCR sequencing analysis and novel target identification (oncogene antigens, private neoantigens, ncRNAs, cryptic peptides).

---

Takeda Oncology (Cambridge, MA)

Principal Scientist: Discovery Translational Science and Computational Oncology
October 2017 – December 2021

• Directed Computational Oncology support for multiple cell therapy discovery projects.
• Collaborated closely with cell therapy teams to design and analyze preclinical translational experiments, informing clinical strategy.
• Led pilot evaluations of spatial transcriptomics platforms, informing technology adoption across Discovery.
• Managed three direct reports, fostering skill development and productivity improvements.
• Spearheaded optimization of cell therapy combinations involving armoring, target selection, platform delivery, and patient population stratification (Project “Square”).
• Identified novel cell therapy targets with an emphasis on clinical relevance and competitive landscape.
• Played key role in interview panels and onboarding, enhancing team cohesion and technical ramp-up.
• Integrated public and proprietary genomic and proteomic datasets to identify tumor-associated antigens uniquely overexpressed in lung cancer.
• Supported GI-DDU scientists with RNA-Seq and scRNA-Seq dataset analyses and external therapeutic platform designs.
• Developed, validated, and disseminated single-cell RNA-seq pipelines, enabling reproducible data analysis across platforms.
• Founded and led internal working group focused on scRNA-seq technology awareness and best practices.
• Revitalized scientific computing environments, facilitating cloud migration and elastic resource scaling.
• Oversaw NGS outsourcing operations for multiple teams, ensuring cost efficiency and data quality.
• Core team member piloting Apple MacBooks deployment across Discovery, supporting IT transition.

---

Ohana Biosciences/VL34 Inc (Cambridge, MA)

Senior Scientist, Computational Biology
October 2016 – October 2017

• Provided foundational scientific analysis culminating in publication in Science.
• Designed and implemented AWS-based collaborative computing environments for large-scale genomics.
• Performed exome sequencing of key commercial dairy bulls to identify novel tissue-specific protein isoforms.
• Developed scalable pipelines for transcript reconstruction and whole chromosome haplotype phasing across species.
• Led multi-species genomic analyses interpreting whole genome, exome, and single-cell sequencing.
• Assembled tissue- and cell-type-specific transcriptomes integrating publicly available datasets.
• Integrated RNA-Seq data from >15 distinct bovine tissues to generate comprehensive gene and transcript annotations.
• Fostered relationships with dairy farmers innovating in cattle genetics, integrating phenotype and genotype datasets.
• Developed candidate evaluation portals integrating multi-omic data to optimize candidate selection efficiency.

---

Takeda Oncology (Cambridge, MA)

Scientist 2, Computational Biologist
October 2014 – October 2016

• Advanced oncology small molecule drug development using HTS data from public and in-house sources.
• Conducted comprehensive bioinformatic assessments covering expression profiles, indication suitability, mutations, and other omic features.
• Identified early translational strategies applying tool molecule responses across hundreds of cell lines.
• Selected translationally appropriate in vitro and in vivo models for target validation.
• Conducted genome-wide screening for antibody drug conjugate targets.
• Analyzed treatment-induced differential gene expression in models.
• Enhanced department-wide computational biology capabilities with command-line and cloud computing adoption.

---

Voyager Therapeutics (Cambridge, MA)

Consultant, Computational Biology Vector Design Group
June 2014 – October 2014

• Built AWS cloud-based computational biology HTS analysis environment.
• Directed AAV vector design leveraging public small RNA-Seq datasets.
• Analyzed and interpreted company-generated HTS data.

---

Graduate School (Worcester, MA)

Graduate Student
See Education for details. September 2007 – June 2014

---

Genzyme (Framingham, MA)

Research Associate, Bioanalytical Development Group
April 2004 – September 2007

• Developed GLP/GMP biochemical assays for recombinant protein clinical trials.
• Engineered HPLC assays for protein glycosylation quantification.
• Developed qPCR assays to monitor viral contamination in production cell lines.

Research Intern, Bioanalytical Development Group
Summer 2003

• Supported preclinical and clinical research via IEF assays and ELISAs.

---

Achillion Pharmaceuticals (New Haven, CT)

Research Internship
Summer 2002

• Conducted ELISAs and cell culture supporting small molecule screening.

---