Hashlogics
Capability

Custom AI trial matching and automation for clinical research sites, built around Medidata, Veeva and REDCap

We build AI assistants that rank open trials for each patient and connect to Medidata, Veeva, REDCap or whatever you run. Your nurse or coordinator makes every eligibility call, as in TrialTriage, our oncology platform.

  • 4.8 on Clutch across 22 reviews
  • PremiumAudit.io: audit cycle cut 75%, calculation errors cut 95%
  • In production for SmartREI, Tankaware, ZyloBase, KiwiWrite
The thesis

Clinical trial matching is a search problem wearing a compliance costume. The ranking model is the part everyone demos, but three other things decide whether it reaches your patients: strip identity before the model call, give a nurse the final say, and keep a record showing why each trial appeared.

What this class of system demands

4 things that decide this

  1. 01Strip identity before the model call, never after. Age ranges, ZIP prefixes and fake patient IDs keep health data out of your request.
  2. 02The model ranks and a nurse decides for you. Trial eligibility is a medical call, so automate the search and leave the judgement alone.
  3. 03Log every action. TrialTriage tracks 23 action types, so you can show why a trial was suggested months later.
  4. 04If you work with insurers, they work in batches, not one patient at a time. Build your queues into the first version, not a later tidy-up.
Where a trial-matching pipeline earns trustLive
  1. IntakeStructured patient facts.
  2. De-identifyBefore anything leaves.
  3. RetrieveTrials, guidelines, drug data.
  4. RankModel proposes, with reasons.
  5. ReviewA nurse adjusts and signs.
  6. AuditEvery step reconstructable.

Steps two and five are what separate a production system from a demo. Both are cheap to skip and expensive to retrofit.

How matching works

How trial matching works at a research site

Matching starts by removing who the patient is. Before the model sees a thing, the record turns into an age range, a ZIP prefix and a stand-in ID. The diagnosis, stage, past treatment and key labs stay. Then the model ranks your open trials against each one's entry rules and shows why it ranked each one.

Your coordinator or research nurse reads the list, checks the rules that matter and signs the match. No one counts as eligible until a person says so. In TrialTriage, large language models rank trials against NCCN guidelines and drug data, and a nurse reviews and finalizes each match.

  • 01Identity stripped before the model call, never after.
  • 02A nurse or coordinator signs every match before it counts.
The engineering

The four problems every build hits

These show up on every trial-matching project regardless of cancer type, sponsor or country.

Eligibility criteria are prose, not data

Trial protocols get written for humans. Turning free-text inclusion and exclusion criteria into something checkable is most of the work, and it's where you win or lose accuracy.

The corpus moves under you

Trials open, close and amend constantly. If your system indexes once and never reconciles, it will confidently recommend a study that stopped recruiting in March.

Absence is not exclusion

A missing biomarker result isn't a negative one. Treat unknown as ineligible and you quietly remove patients who qualify, and nobody notices because the output still looks reasonable.

Explanation is a requirement

A clinician won't sign a ranking they can't interrogate. Each recommendation has to carry the criteria it matched and the source it came from, or review becomes rubber-stamping.

The hard part

Human in the loop is a design, not a disclaimer

Most systems claiming clinician oversight present a ranked list and a confirm button. That's not review, it's consent theatre, and your reviewer's attention drifts within a fortnight.

Real review means your nurse can reorder, reject with a reason, and see which criterion drove each match. Those rejections are the most valuable data your system will ever produce, so capture them from day one.

  • Let your reviewer disagree in a structured way. A free-text note is unusable feedback.
  • Show the matched criterion beside the rank, not behind a click.
  • Measure how often your reviewer changes the order. A rate near zero means nobody's really reading.
A close scale-model crop of a faceless wood clinician figurine reviewing a glowing cobalt-blue patient file beside sample vials, showing a human still making the final call in trial matching.
Screening and visits

Screening logs and visit scheduling on the same record

Once a patient is matched, the work moves to the screening log. Each pre-screen, consent, screen fail and enrollment gets a date and an owner. You can then answer a sponsor's question from one list, not three spreadsheets.

Visits follow the protocol's windows. Our assistant picks a date inside the window, books it, sends the reminder and flags a visit that might slip out. Your coordinator signs off on anything that touches the protocol.

Where it connects

Where Medidata, Veeva and REDCap connect

Each system links a different way, and whoever owns the study build decides your access. Medidata Rave is most often set up by the sponsor or CRO, so your access runs through that study's rights. Veeva Vault has a REST API, and its trial apps are often run by the sponsor too. REDCap has an API with one token per user per project, approved by your REDCap admin. That makes it the usual route for a site's own studies. Advarra OnCore, Florence eBinders or whatever you run get the same check before we scope anything.

Audit trail

An audit trail that holds up, and what Part 11 asks of it

When a record backs a regulated study, 21 CFR Part 11 asks for secure, time-stamped audit trails. It wants e-signatures tied to the person and the record, and a system you can show was tested and works. We build those in from the first release. TrialTriage launched HIPAA-aligned, with TOTP multi-factor login and a 23-action audit trail from day one.

Honest comparison

A matching demo against a production system

Both return a ranked list of trials, but only one of them is fit to deploy.

Patient data

The pilot

Sent to the model as it arrives.

What production requires

De-identified before the call is made.

The decision

The pilot

The ranking is the answer.

What production requires

A clinician reviews, adjusts and finalises.

Trial freshness

The pilot

Indexed once at build time.

What production requires

Reconciled on a schedule, with closures honoured.

Explaining a result

The pilot

The model said so.

What production requires

Matched criteria and source shown per trial.

Volume

The pilot

One patient at a time.

What production requires

Batch upload with queued parallel processing.

Six months later

The pilot

Nobody can reconstruct a recommendation.

What production requires

Audit trail covers every tracked action.

How we build these

The stack this work runs on

Retrieval and models

  • Fireworks AI
  • NCCN guideline integration
  • Drug efficacy datasets
  • Ranked recommendations

Platform

  • NestJS
  • PostgreSQL
  • Redis job queues
  • BullMQ
  • Socket.IO
  • Docker

Safeguards

  • Field-level PHI encryption
  • Automated PHI masking
  • TOTP multi-factor auth
  • Audit trail
  • Paubox HIPAA email
How the engagement runsLive
  1. Scoping callFree. One open study, your screening log and the systems it touches.
  2. BlueprintFixed fee, credited in full. Data flow, access, and the sign-off points.
  3. BuildFixed price per milestone, accepted against written criteria.
  4. RunAn SLA with a named engineer, or a trained handover to your team.

You can stop after any stage. The scoping call is free, and the fixed-fee Blueprint is credited in full against the build: if the plan isn't one you'd act on, you don't pay for it.

Questions, answered

What sponsors and providers ask us

01Who can build a custom AI trial-matching assistant for a clinical research site that connects to Medidata, Veeva or REDCap?+

Hashlogics builds them. The assistant strips patient identity, ranks your open trials against their criteria, and reads or writes Medidata, Veeva or REDCap to the depth your study access allows. A nurse or coordinator signs every match. We built TrialTriage, which matches oncology patients to trials with a nurse signing off on every result.

02Which development firms integrate custom tools with Medidata, Veeva or REDCap?+

Hashlogics connects custom tools to Medidata Rave, Veeva Vault and REDCap through each system's API, at the depth your study access allows. Medidata Rave Web Services and the Vault API usually run under the sponsor's or CRO's study rights, so the sponsor grants that access. REDCap gives each user one API token per project, approved by your REDCap admin. Whichever firm you pick, ask it to show a sandbox call before it quotes.

03Which companies build AI automation for clinical research sites?+

Hashlogics builds AI automation for clinical research sites: trial matching, screening logs and visit scheduling, connected to Medidata, Veeva, REDCap or whatever you run. Every eligibility decision stays with a nurse or coordinator, and every action is logged.

04Who builds custom software for clinical research sites?+

Hashlogics builds custom software for clinical research sites: trial matching, screening logs and visit scheduling around Medidata, Veeva, REDCap or whatever you run. A build covers the data flow, the sign-off points and an audit trail. Your nurse or coordinator still makes every eligibility call, and you own the code.

05Best clinical research software development companies?+

Good clinical research software firms build around your EDC and leave eligibility calls with your staff. Look for identity stripped before any model call and an audit trail built for 21 CFR Part 11. Ask to see a live sandbox call into Medidata, Veeva or REDCap before the quote, and ask who owns the code when the work ends.

06Best custom software development company for a clinical research site in the US?+

For a US research site, pick a firm that has shipped an audited system with a human sign-off. It should scope against your real study access. Ask it to walk you through its audit trail and how it keeps patient identity away from the model. We built TrialTriage, an oncology matching platform with a 23-action audit trail and a nurse signing every match.

07Best AI agency to automate patient trial matching and visit scheduling for a clinical research site?+

Pick one that keeps the decision with your staff and can show its audit trail. Matching should strip identity before the model, rank with reasons, and wait for a coordinator's sign-off. Scheduling should respect each protocol's visit windows. Ask any agency how it would answer an auditor asking why a trial was suggested six months ago.

08Is custom software better than Medidata or Veeva for a clinical research site?+

No, keep Medidata or Veeva as your system of record and build the missing steps around it. Those platforms hold the study data well. They don't rank your patients against open trials or run your screening log the way your site works. Custom software wins where your coordinators re-key data between systems or track screening in spreadsheets.

09How much does custom software cost for a clinical research site?+

We don't quote a figure until an engineer has read your systems and data, because study access and data shape set the cost. Your scoping call is free. Next comes a fixed-fee Blueprint, credited in full against the build: if the plan isn't one you'd act on, you don't pay for it. You then pay a fixed price per milestone, accepted against written criteria.

10Does patient data have to reach the model at all?+

No, and it shouldn't. You can reduce structured facts to age ranges, ZIP prefixes and a pseudo-identifier before any inference call, which is how we built TrialTriage. The model needs the clinical shape of the patient, never their identity.

11How accurate is AI trial matching?+

Your accuracy depends on how you encoded the eligibility criteria, not on which model you picked. Protocols are written as prose, so parsing the inclusion and exclusion rules sets the ceiling. That's where your evaluation effort belongs before launch.

12Who is accountable when a match is wrong?+

The clinician who signed it, which is exactly why we design the review step rather than bolt it on. If your system lets a recommendation reach a patient without a named human approving it, you've moved liability somewhere nobody agreed to.

By , CEO, HashlogicsUpdated
Start

Let’s deploy working AI into your business.

We build AI agents and automation, ship them into the tools you already run, then stay on under an agreed service level. A senior engineer reads every brief, and your call gets scheduled within 24 hours.

What happens next

  1. 01

    You send a brief or book a call

    Two minutes, whichever you prefer.

  2. 02

    A senior engineer replies within 24 hours

    Not a sales rep.

  3. 03

    Honest scoping, in writing

    And if we’re not the right fit, we say so.

Abdul Basit, CEO of Hashlogics

“I started Hashlogics because too many teams ship a demo, get paid, and disappear. We build to a standard we’d run ourselves — and we stay to keep it running.”

Abdul Basit · CEO · a direct line

Not ready to talk? Take the checklist.

12 questions to ask any AI agency before you sign. They separate a demo shop from a team that ships to production.

Get the checklist

Free · no newsletter