Dr Raminderpal Singh

Dr Raminderpal Singh

AI Engineer (Life Sciences)

Global Head of AI/GenAI Practice · 20/15 Visioneers
Founder · HitchhikersAI

What I am building

I'm treating this stage of my career as my post‑doc: building scientific reasoning agents on open‑source LLMs, focused on pre‑clinical drug discovery. Andreessen Horowitz’s November 2026 report, Big Pharma Can Rival Big Software, names scientific reasoning agents as one of five technology categories in the AI opportunity for drug discovery: LLMs applied to science that reason over literature and datasets at scale to find overlooked opportunities. I'm interested in work with large impact, the kind that makes a significant difference to lead scientists. I draw on my systems engineering experience and write executable specifications, and I build apps every day with Claude as my development partner.

Focus: GLP‑1 programs

Across my coding projects, my focus is GLP‑1 programs, where pre‑clinical understanding of the diseases and the compound pipelines are still evolving, so there is much to learn. The projects build on one another: the reasoning core of project fred runs inside Compound Insights and the SIM Framework, ScienceClaw builds a GLP‑1 clinical trials knowledge base that the SIM Framework uses, and all of them feed caffeinated.bio, the project where I am aiming for the largest impact. ScienceClaw began as a side project and shaped the thinking behind project fred.

Working within constraints

All of them are compute‑ and information‑constrained systems by definition: they run within tight compute budgets and reason over incomplete evidence. My engineering experience applies directly to building under those constraints, and pre‑clinical drug discovery works under the same ones, with scarce data and decisions made before the evidence is complete.

How I use LLMs for science

For accurate scientific reasoning, I use LLMs to reason over the evidence I supply, not to recall science from their training data, whether intentionally or by accident. Recalled knowledge has no source, may be out of date, and cannot be checked, so every claim must trace to a supplied citation and is marked when it does not. I currently use Qwen3.8‑27B, an open‑weights model that runs locally and, in my testing, performs well at this task. For the scientific output to be effective, two things must be tuned: the harness, the workflow that manages the model, must be tuned to the model in use, and the model and the specific science, whether a compound, an indication or a pathway, must be tuned together within the system.

HitchhikersAI

300+
community members

A non‑profit grass‑roots community accelerating the adoption of AI/ML and data in scientific R&D — starting with drug discovery & development. Members include bench scientists, data scientists, mathematicians, business owners, executives, and academics — all focused on fixing the disconnect between AI/ML/GenAI and its practical application in the lab.

AI in Drug Discovery

Regular column in Drug Target Review exploring the real‑world application of AI, ML, and generative AI in drug discovery — cutting through the hype to examine what actually works, what doesn't, and what the industry needs to do differently.

LLM in Life Sciences News Tracker

A curated tracker covering AI scientists, autonomous discovery systems, and infrastructure across pharma and biotech — from funding rounds and platform launches to partnerships and regulatory developments. Searchable and filterable by category. Updated weekly.

AI Coding: A Practical Approach for Scientists

Talk given at BAGIM on September 17, 2026: how a scientist who doesn't code can design, build, test and ship data science applications, with Compound Insights as the worked case study. The full talk is on the page — slides, diagrams and the presenter notes for all 23 slides, including the failure patterns and the checks that catch them.

Decision Support Without Agentic Engineering

Positioning paper, preprint — one pinned reasoning engine and a human who decides

Right about process, wrong about architecture — I keep agentic engineering’s external checks and human judgment, and reject many agents converging on an answer where a scientist or a committee must decide.

Two example applications — How Compound Insights and the SIM Framework are built and structured, and why the conventional coding-agent build was not taken.

September 2026 — ↓ download the PDF

SLAS Discovery Special Issue

Guest Editor — Special Issue: Real‑Life Case Studies in AI‑Driven Drug Discovery Workflows

AI is everywhere there is data — data science, bioinformatics, cheminformatics, ML, LLMs. One community, many labels.

Deploying it is now routine — academics, vendors and labs have made AI mainstream.

Share what happened, failures included — so others can learn.

The stories get heard — an active readership of practitioners and academics.

Co-edited with Srijit Seal, PhD. Submissions until December 15, 2026 — ↗ read the call for papers

I’ve spent 20 years moving between technical, commercial, and leadership roles across life sciences, semiconductors, and data infrastructure. That range matters — because the AI adoption challenge in scientific R&D isn’t purely technical. It sits at the intersection of engineering discipline, domain expertise, and the ability to navigate enterprise-scale organisations.

20/15 Visioneers

20/15 Visioneers

Global Head of AI/GenAI Practice · Current (2 years)

Partnered with John Conway (Founder & Chief Visioneer) to address AI adoption as the interconnected challenge it actually is — combining AI engineering, LLM agents, and in-silico software with organisational change, culture transformation, and FAIR data strategy.

incubate.bio

Founder & CEO · 3 years

Founded and led a company building computational platforms for drug discovery. The core product — ALaSCA — applies Pearlian causal inference to multi-omics data. Four bioRxiv preprints: DDR resistance in cancer (2024 ↗), pathway simulation in Type 1 Diabetes (2023 ↗), causal inference in Alzheimer’s (2022 ↗), and ML target prioritisation in aging (2022 ↗).

Eagle Genomics

VP / Head of Microbiome Division · VP Business Development · 3 years

Led market development and product strategy across epigenetics, microbiology, multi-omics, and real-world evidence. The company’s top seller, regularly closing multi-year six-figure solution deals with blue-chip life sciences and CPG customers worldwide.

IBM Research

Business Development Executive · 7 years

Led the Watson genomics programme in partnership with the New York Genome Center, reporting directly to an IBM Senior Vice President. Closed multi-million dollar agreements in healthcare & life sciences, including complex IP licensing and partnership contracts.

IBM Semiconductor Group

Senior Manager & Senior Member of Technical Staff · 6 years

Program Director for Operations Research at IBM’s 300mm Fishkill fab. Led a cross-functional team across multiple IBM organisations. Developed and deployed a predictive analytics platform with IBM Research, saving $10M+. Awarded 12 patents during this period.

Top 13 Influencers in the Semiconductor Industry

EETimes, 2003

Signal Integrity Effects in Custom IC and ASIC Designs

Book · Author

Silicon Germanium: Technology, Modeling, and Design

Book · Author

For the full picture — including patents, earlier publications, education, and additional roles — see my LinkedIn profile ↗