Research Operations Partner

The Hacking Games About Deep Signal Deep Signal is a transcript-based psychological risk screening engine for high-trust roles. Rather than relying on self-report questionnaires — which are fast and cheap, but easy for a motivated individual to game — Deep Signal analyses interview transcripts the way an experienced clinical psychologist would: reading how someone talks about themselves and others to surface patterns that are very difficult to fake. The result is a classification and diagnostic assessment that gives organisations clinical-depth insight into a candidate's psychological risk profile at a fraction of the cost and speed of traditional clinical assessment.  The technology sits at the intersection of clinical psychology, applied AI, and high-stakes vetting — with early interest from cybersecurity hiring, defence and government vetting, and financial services conduct-risk screening. We're an early-stage, f The team Deep Signal is a small team by design at this stage, operating as an intrinsic part of The Hacking Games:  * Mark Loftus — clinical psychologist and the architect of Deep Signal, currently the primary source of clinical judgement on the project. * A retained technical/AI advisor, running the software development of the scoring pipeline and infrastructure. * A search underway for a senior clinical/forensic psychologist to build clinical depth alongside Mark. * Deep Signal sits within the wider organisation (The Hacking Games, currently a team of 20 people) with an existing large-scale cyber-aptitude and talent platform (Haptai), which feeds candidate data into Deep Signal's calibration work. This role would be the first dedicated hire focused specifically on the discipline and rigour behind the scoring methodology and pipeline itself.  How this role fits in This role exists to take on: rigorous testing, clean record-keeping, and clear diagnostic evidence — freeing Mark to spend his time on the clinical and strategic judgement calls that genuinely need his expertise, while you make sure that judgement is being tested properly rather than taken on faith. You'll report directly to Mark and work closely with our technical advisor on the underlying pipeline infrastructure.  What You Will Do Testing and calibration discipline * Design and run structured tests every time we change a scoring prompt, marker, or protocol — using a fixed reference set of transcripts, so we can say with confidence what changed and why, rather than relying on impressions. * Keep a clean, versioned record linking every output to the exact model, prompt, and protocol version that produced it. Diagnostics and analysis * Analyse which markers and signals are actually driving the model's classifications, and flag where something expected isn't showing up as expected. * Track and investigate unexpected shifts in output — for example, if a change we thought was minor ends up changing someone's diagnostic classification, you'll be the one who traces why. * Support the statistical and regression work as we build up a larger set of real interview transcripts over the coming months, helping move the model from expert-judgement-calibrated to properly data-calibrated. Documentation * Build the habit — for yourself and eventually the wider team — of documenting reasoning and decisions as they're made, not reconstructing them later from memory. * Help turn what is currently a set of working scripts and informal knowledge into something a new team member could actually pick up and understand. A standing responsibility to flag, not to decide You will regularly be looking at results that raise questions — an unexpected classification, a marker behaving oddly, a pattern that doesn't quite make sense. Your job is to surface these clearly and rigorously, with evidence, and push back when something looks wrong. The clinical interpretation and the calls that follow from it stay with Mark (and later, our senior clinical hire) — this is a deliberate design decision, not a reflection of trust, and it's discussed openly with everyone on the team.  What we're looking for This is not a conventional data science role, and it's not a clinical psychology role either — it sits in between, and we're open-minded about which side of that line you come from, provided both are genuinely present. What You'll Bring * A strong quantitative/analytical background — comfortable with statistical concepts (regression, feature importance), and either already experienced with, or a fast learner of, working with AI/LLM prompts and outputs. * Real intellectual curiosity about psychology — enough to understand what a concept like "mentalising" or a specific marker is meant to capture, and why its presence or absence might mean different things in different cases. Formal clinical training is not required, and we're not looking for someone who intends to become a practising psychologist — we're looking for someone who finds this subject matter genuinely interesting to think about rigorously. * A natural inclination toward rigour and process for its own sake — designing a clean test, keeping tidy records, and documenting things properly should sound satisfying to you, not like a chore standing between you and the "real" work. * Comfort working in an early-stage, still-forming environment, where a lot of what you'll be doing is helping establish process that doesn't fully exist yet, rather than following an established playbook. You might come from * A research assistant or lab manager role attached to a computational or quantitative psychology research group. * A postgraduate or recent research background in computational psychiatry, psychometrics, or applied natural language processing, with a specific interest in psychological measurement rather than machine learning in the abstract. * Another applied research or data role where you were responsible for the rigour behind someone else's substantive decisions, rather than making those decisions yourself. What success looks like in the first few months * You've built and are running a structured testing process for pipeline changes, and Mark is using your test results — not just his own read — to decide whether a change is an improvement. * There's a running, organised record of what's changed in the scoring pipeline and why, that anyone on the team could pick up and understand. * You've independently spotted and flagged and suggested a resolution for at least one thing that is genuinely material to the success of the Deep Signal engine. Future prospects This is a genuinely early-stage role on a product we believe has significant potential, in a small and growing team. As the calibration and validation work matures over the coming year, this role has real scope to grow — potentially into ownership of the technical/data side of Deep Signal's product development as the team scales, working alongside (rather than under) whoever leads the clinical side. Where exactly this role goes will depend on how Deep Signal grows and where your own interests take you, and that's a conversation we're genuinely open to having as we go, not a fixed script. Apply Email hr@thg-v.com with: