PDAOAI – Oncotelic

BIO International Convention 2026 · Booth #1525

It's not a search problem. It's a signal problem.

PDAOAI™ is Oncotelic's Knowledge Discovery platform, powered by Structural Intelligence — it folds 28 million papers into structure and finds the root driver, not proximity. At BIO 2026, see it in two modes: Knowledge Discovery for drug discovery, and PDAOAI Robotics — PUR-E and LIM-E for AI-powered GMP manufacturing. Same engine. Two corpora. One booth.

01 · DRUG DISCOVERY

Knowledge Discovery

Structural Intelligence — it finds the root driver, not the nearest match. From 28 million PubMed papers, surface the upstream gene, mechanism, or theme — the few things everything else moves around. Vector search returns proximity; we return the signal.

Proven the hard way: designed around the Sapu Nano pipeline — the engine that took two drugs, SAPU 003 and SAPU 006, from concept to first-in-human in 24 months.

Proof: Asked PDAOAI what drives a cancer. Got YAP1. Held up against patient data it had never seen.

Partner: Built on Qdrant — on stage together at Qdrant Vector Space Day 2026.

02 · GMP MANUFACTURING

PDAOAI Robotics

The same engine, pointed at your SOPs and batch records — riding on two autonomous robots: PUR-E in the cleanroom, writing the electronic batch record hands-free as the operator dictates through Kebbi, and LIM-E moving materials in the corridors.

Partner: TechForce Robotics (OTCQB: NGTF). Phase 1 LIM-E deployment complete at Oncotelic.

Built on Qdrant. We run 28 million vectors in production on its hybrid search, payload filtering, HNSW indexing, quantization, and TurboQuant.

Dates: June 22 – 25, 2026
Location: San Diego Convention Center
Find us: Booth #1525, CLS Pavilion

Visit the joint exhibit experience → Request a meeting

● 28 million abstracts, live on Qdrant Cloud

The totality of biomedical knowledge, put to work.

PDAOAI™ is Oncotelic's knowledge discovery platform — proven the hard way. It was designed around the Sapu Nano pipeline — the engine that took two drugs, SAPU 003 and SAPU 006, from concept to first-in-human in 24 months. It reads the full public PubMed corpus — about 28 million abstracts. LLMs and vector search fall apart on biology — the signal drowns in the noise and "closest" returns proximity, not cause; PDAOAI's proprietary clustering, folding, and shell-sampling methods reshape the hyperdimensional space and sample a bounded shell of the manifold — clearing the ghosts, the phantom matches that look related but aren't — to surface the root driver. This is knowledge discovery, not information retrieval — Structural Intelligence that returns insight, auditable to the source and embeddable into the systems where research and manufacturing actually happen.

Drug discovery → PDAOAI robotics →

~28M

PubMed abstracts indexed · refreshed daily

<2 yrs

discovery-to-IND, two programs

2 INDs

SAPU 003 & SAPU 006 filed

Live

production-scale on Qdrant

Built on Qdrant. We run 28 million vectors in production on its hybrid search, payload filtering, HNSW indexing, quantization, and TurboQuant.

One platform, two applications

From the literature to the lab floor

One engine. Two jobs. PDAOAI finds the root driver in the literature — then runs that same knowledge layer on the GMP floor.

🧬

Find the root driver

Vector search hands you the crowd. PDAOAI hands you the cause — the upstream gene everything else moves around.

See how it finds YAP1 →

🤖

Run it on the GMP floor

The same engine, riding two robots into the cleanroom — writing the batch record as the work happens, right the first time.

Meet PUR-E & LIM-E →

Theme 1 · Knowledge Discovery

Find the root driver the crowd can't see.

We pointed an AI at 28 million papers and asked one question: what drives a cancer? It returned one gene — YAP1 — and it held up against patient data it had never seen. That's Structural Intelligence: it finds the root driver, not the nearest match.

Not better search. Not bigger AI. It's not a search problem — it's a signal problem.

Today's AI can't do this. A bigger context window holds more text, not more understanding — the signal drowns in the noise — and vector search returns proximity, not cause. PDAOAI's proprietary methods — clustering, manifold folding, and shell sampling — reshape the hyperdimensional space so related concepts cluster and unrelated ones pull apart, then surface the root driver from a bounded shell. Knowledge discovery, not information retrieval.

Ingest → Fold → Sample → Discover

Ingest

Index ~28M PubMed abstracts and proprietary corpora — the totality of biomedical knowledge, refreshed daily.

Fold

Fold the hyperdimensional space so related concepts cluster tightly and unrelated ones pull apart.

Sample

Sample a bounded shell of the manifold, clearing the ghosts — the phantom matches that look related but aren't.

Discover

Surface the root driver: the upstream gene, mechanism, or theme everything else moves around.

Fold the space → clear the ghosts → the root driver glows. Search finds the needle. We find the magnet — the thing every needle points to.

What folding buys you

Smaller search spaceTighter, separable regions to draw from.
Lower computeLess to scan for the same recall.
Better LLM qualityCleaner evidence in, fewer hallucinations out.
Quality-awareIt even separates evidence strength — strong from weak — not just relevance.

Protected by a patent family

PDAOAI's clustering method is the subject of an international patent application under the Patent Cooperation Treaty (PCT/US2024/057972). It drew a favorable International Search Report and Written Opinion — all claims assessed as novel, inventive, and industrially applicable. Manifold folding and shell sampling are covered by further applications in the same patent family.

Pharma — the proof

Using only published papers, PDAOAI named YAP1 as the root driver of a cancer — the gene at the structural center, not the most-studied one. It held up against survival, single-cell, and independent data it never saw. Dr. Vuong Trieu produced it on the live platform, and it reproduces.

And we've run the whole play: two drugs, SAPU 003 and SAPU 006, from concept to first-in-human in 24 months — plus the GMP plant to make them, built from scratch through Sapu Nano. Most AI drug startups begin with molecules; we begin with the biology that drives the disease.

It moves target validation from a late-stage clinical failure to a day-one computational check.

"Qdrant's primitives — hybrid search, payload filtering, HNSW indexing, quantization — give us the production layer to run 28 million vectors at a cost structure that lets us advance more programs." — Saran Saund, Chief Business Officer, Oncotelic

"What Oncotelic has built on top of our primitives — manifold folding for biomedical literature — is one of the most ambitious applications of vector search in healthcare we have seen." — André Zayarni, Chief Executive Officer, Qdrant

What this enables

Root driver, not proximity.Find the upstream gene, mechanism, or theme — not the most-cited match.
Whole-corpus recall.Evidence from all of PubMed, refreshed daily.
Day-one validation.Move target validation from Phase II failure to a computational check.
Auditable.Every answer traces back to the exact sources, ranked by structure.
Multi-LLM, online or on-prem.Run it where your data lives.
Same engine, swap the corpus.Drug discovery, clinical trials, manufacturing SOPs — in daily production use.

Search finds the needle. We find the magnet — the thing every needle points to. — PDAOAI

"Building the DNA of Search" — Qdrant Vector Space Day 2026

PDAOAI on stage with Qdrant — Product & Engineering keynote, June 11, 2026, San Francisco, with Qdrant's Head of Product, Bastian Hofmann.

Read the announcement →

Point it at your corpus.

PubMed, your SOPs, your batch records — same engine, your data, online or on-prem.

Request a demo →

Theme 2 · PDAOAI Robotics

Knowledge that moves to where the work is

Indexing the literature is step one. Step two is putting it to use. The batch record is the heart of GMP — PDAOAI runs it. Under a joint development agreement with TechForce Robotics, that capability rides on two autonomous robots inside the facility: PUR-E in the cleanroom, writing the batch record hands-free through its Kebbi voice interface, and LIM-E in the corridors, moving the materials. Phase 1 objectives are complete; both companies are evaluating and co-developing further automation for GMP manufacturing and laboratory workflows.

Two robots, two jobs

Unit 01 · Cleanroom & eBR · Must-have

PUR-E™ — Purity In Motion, with Kebbi

The batch record is the heart of GMP — PUR-E runs it. It pulls the room's particle and sterility-monitor readings straight into the eBR, and lets gowned operators dictate batch parameters hands-free through the Kebbi conversational interface — PDAOAI writing the record as the work happens, right the first time.

LIM-E logistics robot carrying a bulk solvent drum

Unit 02 · Logistics · Efficiency layer

LIM-E™ — Logistics In Motion, Everywhere

Autonomous transport of materials and equipment — trays, consumables, kits, bulk solvent drums, and waste, up to 600 lb per load — running 18–20 hours a day in the non-classified corridors, handing off at the cleanroom pass-through. Built for heavy loads, so qualified staff don't carry them.

Both robots are tested in operating sterile cleanrooms and built for GMP — low-particulate, fully wipe-down construction.

What it returns

Conservative, modeled on a mid-size sterile suite (~60 batches/year). Effects shown as ranges and confirmed per facility.

PUR-E™ · Cleanroom & eBR · Must-have

Batch-record review time	−40–50%
Manual data entry	−60%
Documentation errors	−50%
Batch-release cycle	−40–50%
Subscription	under ~1% of suite throughput

Records-QA workload shifts off the critical path; faster release frees working capital tied up in held inventory. Upside not modeled: an avoided scrapped batch or FDA 483 observation. One-time: eBR/MES validation (21 CFR Part 11).

LIM-E™ · Logistics · Efficiency layer

Material-handler labor	offsets ~1–1.5 porter FTE
Payload	up to 600 lb per load
Uptime	18–20 hrs/day · >1 shift
Subscription	under ~1% of suite throughput

An efficiency add-on, not a must-have. 36-month agreement, cancel on 30-day notice. Fewer human trips in classified space — lower contamination risk; staff kept in-suite on value-added work.

See it live at BIO 2026 — Booth #1525, CLS Pavilion

PDAOAI + PUR-E & LIM-E demonstrated on the BIO floor June 22–25, San Diego. Hands-free eBR through Kebbi, AI-assisted documentation, intelligent robotic integration.

Visit the joint exhibit experience →

Two AI layers, kept distinct

A robot on a regulated floor runs two different kinds of AI. PDAOAI is the knowledge layer; TechForce's onboard stack drives the robot. Each does a different job.

Oncotelic · Knowledge

PDAOAI — the AI knowledge layer

Indexes the totality of biomedical knowledge and proprietary corpora, and embeds it into automated workflows — supporting research, biomarker discovery, regulatory processes, and compliance across pharmaceutical development and manufacturing.

TechForce · Onboard

Sensing & navigation

LIM-E runs LiDAR-based SLAM navigation, 3D depth and vision sensing, and adaptive AI that plans movement continuously and reroutes around people and obstacles.

TechForce · Onboard

Facility integration

Robots call and ride commercial elevators, open automated and badge-controlled doors, and dynamically reroute when hallways are busy — fitting existing workflows without operational redesign.

Expanding Autonomous Operations into Pharmaceutical Manufacturing

TechForce's autonomous platform has been successfully deployed across hospitality, entertainment, and large-scale commercial environments. The LIM-E deployment represents its first operational expansion into pharmaceutical manufacturing and laboratory operations, where automation must meet the rigorous requirements of GMP compliance, data integrity, and process control.

Combined with PDAOAI's domain-specific knowledge layer, the integrated platform is designed to support scalable pharmaceutical operations by improving workflow efficiency, reducing dependence on manual processes, enhancing documentation accuracy, and strengthening compliance across research, development, and manufacturing environments.

"We are moving from data to real-world application. By embedding the totality of scientific knowledge into robotics, we can look to transform how drugs are developed and produced — not just for us, but for the broader industry." — Dr. Vuong Trieu, Chairman & CEO, Oncotelic

Where it stands

Phase 1 complete.LIM-E logistics deployment under the joint development agreement.
Phase 2 underway (June 2026).PUR-E deployed alongside LIM-E in an OEB-5 sterile injectable cGMP facility.
Knowledge embedded.~28M abstracts integrated for real-time use in the robotics workflow.
Next.eBR integration, multi-floor navigation, and further GMP workflow automation.

Newsroom