A drug discovery program can ask a young company to wait ten years. A young company may have two. That mismatch helps explain BIOPTIC, an AI business whose current proposition begins with a question that can be answered much sooner: which drug opportunity deserves your attention?
- AI agents assemble pharma evidence; specialists review the findings.
- Customers commission asset scouting, diligence, competitor research and launch planning.
- Its molecular-screening roots produced published research. Its commercial tools tackle the homework around a deal.
In an August 2026 interview with Pharmvestnik, co-founder and CEO Andrey Doronichev described learning to turn down preclinical projects. Their planning horizon, he said, was incompatible with startup life. Of more than fifteen pharma projects, roughly ten had produced patents or papers; others had exhausted their budgets or delivered insufficient results.
There is a useful business story in that admission. An algorithm can work while the project around it moves too slowly. BIOPTIC’s response has been to carry its scientific research experience toward the commercial decisions surrounding medicines.
First, the market disappeared
The company began somewhere improbable: copied NFTs. Optic, founded by Doronichev, Roman Doronin and Vlad Vinogradov, raised $11 million in seed funding announced in July 2022. Kleiner Perkins and Pantera Capital led the round. OpenSea was an early customer for technology that recognized reused digital content.
Then the crypto market deteriorated. In a later Oninvest interview, Doronichev connected the collapse of FTX with the search for another business. Silicon Valley Bank’s failure added a separate shock: about $10 million of the company’s cash had been held there. That was money temporarily caught in a bank failure, rather than a disclosed loss of the entire balance.
A conversation with cancer researcher Artem Cherkasov helped point the team toward molecules, Doronichev told Pharmvestnik. Could the technology search chemical space? The domain changed dramatically. The attraction of finding useful matches in enormous collections survived.
The molecule still had to meet the lab
BIOPTIC’s B1 model searches for molecules using information from known ligands, compounds that bind to a target. Its public methods paper appeared on arXiv in June 2024. Later work included a 2025 JCIM publication involving LRRK2, a target studied in Parkinson’s disease.
The more revealing example comes from a Stanford-led Friedreich’s ataxia study. According to BIOPTIC’s June 2026 account of the Cell Chemical Biology publication, B1 started from a single known Miro1 ligand and searched 3.2 billion molecules. It returned 62 candidates; 53 were synthesized and tested. Two leading compounds matched or exceeded the reference compound’s performance in patient-derived sensory neurons.
Counts from BIOPTIC’s account of the Stanford-led study. Laboratory findings do not establish efficacy in patients.
Those numbers describe a funnel, with experimental work at its narrow end. Searching billions of possibilities is useful when the resulting shortlist earns a place on the laboratory bench. Patient-derived cells offer evidence about biology; a successful medicine requires much more.
A better brief beats a bigger pile
Today, BIOPTIC leads with research services for pharma teams and investors. A business development group might ask for Phase 2 oncology assets in China that fit licensing criteria. A commercial team might need a comparison of competing programs. A launch team might want to understand treatment pathways across several countries.
The platform divides a question into tasks and sends specialized agents to investigate in parallel. They search trial registries, papers, patents and company disclosures, including Chinese-language material. A knowledge graph connects assets, companies, trials and documents. Reports distinguish supporting evidence from analysis and flag gaps for review.
Pharma professionals check the work before delivery. The leadership roster makes that pairing visible: AI and engineering executives sit alongside scientific diligence specialist Ilya Yasny and commercial diligence specialist Dmitry Kobyzev. The organization’s expertise spans software, scientific interpretation and commercial assessment.
BIOPTIC advertises 24-48 hours for asset scouting and diligence, typically 48-72 hours for an initial competitive-intelligence report, and 72-96 hours for launch research. Monitoring can continue on an agreed schedule. These are service expectations, dependent on the scoped assignment.
That puts the company between familiar alternatives: an internal analyst team, a specialist consultancy and a general AI research tool. Its pitch combines a repeatable software process with a reviewed deliverable. For a licensing team, the practical output is a ranked asset table with current status and supporting records. For an investor, it can be a memo that makes the unanswered questions explicit before a meeting. The value depends on how well that output fits the decision.
Count what the machine misses
A competitive landscape is unforgiving. One medicine can have several names. Two databases can classify the same indication differently. A relevant disclosure can sit behind a licence or appear in another language. A plausible answer can still leave out the asset that changes the deal.
BIOPTIC’s AAAI paper constructed an evaluation corpus from five years of a private biotech fund’s diligence memos. Its competitor-discovery agent achieved 83% recall, compared with 65% for OpenAI Deep Research and 60% for Perplexity Labs in that test.
Company-authored AAAI study. Specific systems and evaluation corpus; these percentages measure recall, not universal accuracy.
The paper also reports an investment-fund case study in which competitive analysis fell from 2.5 days to about three hours. The results support a particular workflow. They leave buyers with a sensible test: does the evaluation resemble the question on their own desk?
Sell a decision people can inspect
The commercial offer is scoped research, with deliverables and timing agreed before work begins. A separate June 2025 collaboration with venture builder Mechanica Partners covered multiple molecules and made Optic eligible for development milestones and tiered royalties. Those economics belong to that agreement, rather than a published price for today’s research service.
“AI for the commercial side of pharma”BIOPTIC’s current positioning
The workflow is worth copying even without the software: define what qualifies, search across names and languages, preserve the evidence, and have someone qualified inspect the exclusions. It depends on accessible records, a clear brief and informed review. Missing disclosures remain missing disclosures. The useful promise is a better prepared decision, delivered while there is still time to make it.