On the morning of October 30, 2025, ICE agents arrived in Woodburn, Oregon. Woodburn is a small city in the Willamette Valley — a majority-Latino agricultural community, home to farmworkers and families. The operation was called Operation Black Rose. Its stated purpose was targeted enforcement against individuals with criminal records.
What actually happened was something different. Before the raid, agents used a new tool called ELITE to survey the area—not to locate a specific person at a specific address, but to map population density. They were looking for what an agent described, under oath, as a “target-rich” area: a zone where enough pins clustered on the map that the probability of a productive sweep was high. The goal wasn’t to find MJMA, the individual eventually detained. It was to find a neighborhood to sweep.
MJMA had no criminal record. She was planning to seek asylum. After her arrest, Innovation Law Lab filed a writ of habeas corpus on her behalf; she was not released for weeks. The suit challenges not just her detention but the method that identified her community as worth raiding in the first place.
The technology that made that possible is Palantir’s ELITE — Enhanced Lead Identification and Targeting for Enforcement. It is a map interface that generates dossiers, assigns confidence scores, and allows ICE agents to draw a circle around a neighborhood and populate it with human targets. The name tells you what it claims to be doing. The court transcripts tell you what it’s actually doing. This research article tells you the difference.
The Sanitized Version
In contract documentation, ICE describes ELITE as “support[ing] the development of an accurate picture of actionable leads … to allow law enforcement to prioritize enforcement actions” — language that positions the tool as an aid to existing legal processes. The implication is efficiency: agents already know who they’re looking for, and ELITE helps them find those people faster. There is always a file. There is always a prior legal finding. The technology just accelerates the workflow.
The name strengthens this promise. “Enhanced Lead Identification and Targeting for Enforcement” strings together administrative terms that each sound like they belong to an orderly legal process. “Lead” implies an existing investigation. “Identification” implies a named subject. “Targeting” implies precision. Put together, the name describes what a warrant looks like in practice. You have a person. You know where they might be. You use technology to find them efficiently.
Palantir, the company that built ELITE, has spent two decades building surveillance infrastructure. The company — co-founded by billionaire Peter Thiel — has contracts with the Pentagon, the CIA, and police departments across the country. Its platforms are designed to integrate vast amounts of data from disparate sources and make that data searchable, analyzable, targetable. We know Palantir built ELITE from procurement records. That same contract modification names ELITE explicitly, tasking Palantir with “configuration and engineering services” for the tool.
In public statements, Palantir argues its objective is to make ICE’s work “more efficient and more accountable.” According to a user guide obtained by 404 Media, ELITE uses address information from the Department of Health and Human Services, US Citizenship and Immigration Services and other sources to “improve capabilities for identifying and prioritizing high-value targets through advanced analytics.” The case they make for its use is that if agents are going to conduct enforcement operations regardless, better data might mean fewer errors, fewer wrong-door raids, fewer cases of mistaken identity.
What’s not to like?
Well… what this framing systematically avoids is how agents choose where to look in the first place. Case management implies a case that already exists. ELITE enables something that precedes case selection: community-level surveying, identifying zones where the probability of finding deportable individuals is high enough to justify deployment. The question is, who are deportable individuals in these datasets and how are such zones identified? This is where ELITE comes in.
The Deportable: What the Pins Actually Are
When an ICE agent draws a circle on the ELITE map, pins populate the area — each representing a specific person. The natural inference is that each pin marks someone who has been investigated, charged, or formally identified as undocumented. That inference is wrong.
A pin represents someone who appears in ICE’s enforcement database — a system built from immigration records, prior enforcement encounters, removal orders, visa overstay flags, and data shared from other federal agencies. Appearing in this database doesn’t mean you’ve been found guilty of anything or even formally investigated. It means your name, at some point, entered a system that ICE now has access to. That entry may reflect actual undocumented status. It may also reflect a data error, a historical record that was never updated, or information shared by an agency that had no enforcement intent when it collected it.
An important feature of ELITE is the “confidence score” which measures two variables: the source of the data (how authoritative the originating institution is) and the recency of the data (how recently it was obtained). What it does not measure is whether the person is actually deportable. It measures address reliability — the system’s confidence that the person is still at the location it has on file. A high confidence score means: we believe this person is at this address. It says nothing about whether the underlying reason they’re in the database is accurate or current.
The National Immigration Law Center has documented that ICE’s databases are riddled with errors — outdated records, duplicate entries, cases where immigration status changed but the database was never updated. A US citizen with a historical immigration record, an asylum seeker whose case is pending, a DACA recipient — all can theoretically generate a pin. This apparatus of enforcement treats presence in the database as a proxy for deportability, then treats address confidence as a proxy for targeting reliability. Neither substitution has been independently audited.
A driving force for such sweeps is the deportation quotas of 3000 a day. ICE agents operate under pressure to meet removal targets. ELITE doesn’t help them find specific individuals against whom warrants have been issued or investigations completed. It helps them identify neighborhoods where the pins cluster densely enough to make sweeps productive towards their quotas. This workflow means, agents don’t check the veracity of a name that appears in ELITE. So, the result is not targeted enforcement of immigration law. It is the targeting of minority communities wholesale.
India’s Aadhaar system demonstrates where this logic leads at scale. Originally built for welfare delivery in 2010, Aadhaar became the de facto identifier for banking, telecom, and voter verification. In Assam, northeastern India, the government linked Aadhaar to the National Register of Citizens (NRC) — a citizenship verification process requiring residents to prove their families lived in India before 1971. 1.9 million people were excluded from the final NRC listpublished in 2019, many of them Bengali Muslims, members of India’s minority community. In 2025, 1.8 million individuals remain in administrative limbo — enrolled in Aadhaar but not formally recognized as citizens, unable to access services, unable to vote. The documentary requirements were designed to be nearly impossible to meet: many residents, particularly those from marginalized communities, don’t have land deeds or birth certificates from 1971. Some families have lived in Assam for generations but lack the specific papers the system demands. The system built for inclusion became an instrument of exclusion. The mechanism is familiar: data collected for one purpose, repurposed for enforcement, with affected communities having no say in the transformation.
What Agents Say Under Oath
I’ve been reading the transcript from a December 2025 court proceeding related to Operation Black Rose. What ICE officials say under oath is substantially different from the official description.
A deportation officer with ICE’s Fugitive Operations Unit, identified as “JB,” described ELITE this way: “One of our apps, it’s called ELITE. And so it tells you how many people are living in this area and what’s the likelihood of them actually being there. It’s basically a map of the United States. It’s kind of like Google Maps.”
Then: “It pulls from all kinds of sources. It’s a newer app that was actually given to us in ICE.”
JB described how ELITE guided the decision about where to deploy for Woodburn: “You’re going to go to a more dense population rather than [...] like, if there’s one pin at a house and the likelihood of them actually living there is like 10% [...] you’re not going to go there.”
This is area selection language. The question isn’t “where is this person I am investigating?” It’s “which neighborhood has enough pins to be worth a sweep?” The operational logic is straightforward: start with an area containing a high volume of pins — each representing anyone with any probability, however remote, of being deportable. Deploy agents to that area. Detain broadly. Some percentage of those detained will turn out to be actually deportable. The rest get sorted out later. The efficiency is in the volume, not the precision.
This happens without warrants. Without individualized suspicion. Without the constitutional protections that typically constrain law enforcement. The justification is stopping crime and removing “illegals” — language that collapses legal distinctions and treats entire communities as suspect. In practice, algorithmic targeting functionally suspends democratic rights for anyone living in a neighborhood the system flags as high-density. ELITE doesn’t ask whether you’ve committed a crime or violated immigration law. It asks whether you live somewhere the pins cluster.
Operation Black Rose demonstrates how this operates in practice. The operation was explicitly designed as a training exercise for ELITE — ICE agents learning the tool while using it on real people. MJMA, one of approximately 30 people detained, had no criminal record. She should never have been detained. But she lived in a target-rich area, which was enough to place her in custody.
Then a second algorithmic system compounded the violation.
MJMA was mis-identified by Mobile Fortify, ICE’s facial recognition database, not once but twice. Mobile Fortify was originally built for border processing, then turned inward to verify immigration status in the field. Research by Joy Buolamwini and others has documented that facial recognition performs significantly worse on people of color. In immigration enforcement, that error rate means a US citizen or lawful resident can be flagged as a deportation candidate because an algorithm misread their face.
And here is the democratic inversion: Mobile Fortify’s determination currently overrides actual legal documentation. We are living through a time when a birth certificate — a legal document establishing citizenship — is less authoritative than a confidence score produced by an opaque algorithm. The black box supersedes the documented right. This is not a technical problem. It is a constitutional one.
The legal architecture enables this. Following an executive order from the Trump administration, data sharing between federal agencies no longer requires the specific legal authorities it previously did. ELITE doesn’t need to get around the law. The law has been adjusted to fit ELITE. What was once a constitutional constraint on government power — requiring warrants, individualized suspicion, documented cause — has been administratively dissolved to accommodate algorithmic efficiency.
Smashed Windows & Assault
Aliya Rahman was on her way to their 39th appointment at a traumatic brain injury rehabilitation center in Minneapolis when ICE stopped them. It was a routine medical visit. Aliya is a disabled Bangladeshi American — both autistic and living with a traumatic brain injury that requires consistent medical care.
She was driving through her neighborhood when she encountered an ICE blockade. No warning. No explanation. Agents had identified the area as high-density — enough pins on the ELITE map to justify a sweep. Aliya wasn’t a target. She was simply there.
An ICE agent smashed her car window and dragged her out. Aliya was never asked for identification. Never told she was under arrest. Never read her rights. Never charged with a crime. She was detained on the spot, four blocks from where George Floyd was murdered in 2020.
The encounter was particularly destabilizing because of Aliya’s autism. Her Congressional testimony describes her brain “fixating on sounds, numbers and patterns” as agents on all sides gave conflicting instructions simultaneously — a sensory overload that made it nearly impossible to process what was happening. Agents laughed as Aliya tried to immobilize her own neck to prevent further injury. When Aliya’s speech began to slur — a sign of serious neurological distress — she asked for a communication navigator. Request denied. She asked for her cane. Agents said no, then mockingly yelled: “Walk, you can do it, walk.”
Aliya repeatedly asked to see a doctor. Request denied until she blacked out. When Aliya woke up, she was in an emergency room being treated for assault.
In a legal system built on innocent until proven guilty, ELITE reverses the equation. The system identifies neighborhoods where marginalized people live, deploys agents to those areas, detains whoever is there, and then — only then — checks whether they have the right person. If they don’t, if they’ve detained an American citizen on their way to a medical appointment, no accountability follows. Just: sorry, wrong person, you’re free to go. Except Aliya wasn’t free to go. She was in a hospital, recovering from what ICE had done to her.
ELITE didn’t identify Aliya as a target. ELITE identified Aliya’s neighborhood. That was enough.
Who Built This, and What Feeds It
Palantir has worked with ICE for over a decade, originally focused on case management for Homeland Security Investigations. That relationship expanded substantially in the second Trump administration. According to public disclosures, Stephen Miller, the chief architect of the Trump administration’s immigration policy, holds a financial stake in Palantir.
The September 2024 contract modification that created ELITE represents a significant escalation. The $30 million ImmigrationOS platform was designed to integrate ICE’s fragmented data systems into a single interface. ELITE sits within that infrastructure as a specific targeting module. Court testimony and reporting confirm that the tool pulls from government databases, benefit programs, and contact information. On December 29, 2025, U.S. District Judge Vince Chhabria ruled that ICE can access Medicaid data from approximately 80 million patients for enforcement planning through Palantir’s platform. Data sharing began January 6, 2026. Medicaid data was collected under an explicit promise it would never be used for law enforcement. That promise is now void.
The technical architecture matters. ELITE doesn’t just combine datasets — it makes them queryable through a geographic interface. An agent can draw a circle on a map and instantly see: how many pins are in this area, what their confidence scores are, whether the density justifies deployment. This kind of geospatial analysis has been used in military contexts for years. Palantir built similar systems for the Pentagon to identify insurgent networks in Iraq and Afghanistan. The methodology is the same: aggregate disparate data sources, weight by reliability, visualize geographic patterns, enable rapid targeting decisions. What changed is the context. The targeting is now domestic. The deployment zones are American neighborhoods.
The technical architecture allows agents to layer different data sources. A Medicaid address from six months ago might receive a lower weight than a motor vehicle registration from last week. An HHS address record might be weighted more heavily than a utility registration. The system performs these calculations automatically, producing a single confidence score that obscures the underlying complexity. An agent looking at the map sees a pin with a score. They don’t see that the score reflects three overlapping address records of varying quality, one of which may be outdated, one of which was collected for healthcare delivery, and one of which came from a database known to have significant error rates.
This is what makes the aggregation so opaque. Individual data sources can be challenged or verified. A Medicaid address can be checked against current residence. A vehicle registration can be cross-referenced with license records. But once these sources are aggregated into a single confidence score displayed on a map interface, that complexity disappears. The pin becomes the person. The score becomes the justification. And the agents deploying to that neighborhood have no practical way to disaggregate the score back into its component data sources to assess reliability.
The score doesn’t reflect legal status or risk. It reflects the system’s confidence that a person is where the system thinks they are. Precision of location is being substituted for precision of targeting.
Palantir knows this is producing problems. A Wired investigation in January 2026 — published after Border Patrol’s killing of Alex Pretti in Minneapolis on January 24, which prompted employee unrest — revealed an internal wiki that acknowledged “increasing reporting around U.S. citizens being swept up in enforcement action and held, as well as reports of racial profiling allegedly applied as pretense for the detention of some U.S. citizens.” The wiki then argues that Palantir’s customers “remain committed to avoiding the unlawful, unnecessary targeting, apprehension, and detention of U.S. citizens wherever and whenever possible.” This is aspiration dressed up as a safety guarantee.
Palantir’s reach extends to the UK’s National Health Service as well. The company secured a contract to manage NHS patient data during the COVID-19 pandemic — a “temporary” arrangement that has since been extended. Privacy advocates have raised concerns that NHS patient records are now accessible to a private surveillance company with deep ties to immigration enforcement and intelligence agencies. The integration of health data with Palantir’s existing law enforcement infrastructure creates precisely the aggregation risk that ELITE demonstrates: data collected for one purpose becoming searchable for another.
The Data Question Underneath
The confidence score’s two inputs — source authority and recency — sound methodologically reasonable. More recent data from more authoritative institutions should produce more reliable location estimates. But this obscures a prior question: which communities are most heavily represented in these data sources to begin with?
Federal benefit programs — Medicaid, SNAP, public housing databases — disproportionately represent lower-income and immigrant communities. The communities most enrolled in these programs are the communities most likely to generate dense pin clusters in ELITE. This is not a neutral data distribution. It reflects decades of policy choices about which populations are most closely monitored by the administrative state.
Consider what this means operationally. A neighborhood where many residents use Medicaid will generate more address records in government databases than a neighborhood where residents use private insurance. A community where residents rely on public housing assistance will have more verified addresses in HHS systems than a community of homeowners. These data sources weren’t created for enforcement purposes. They were created to deliver services. But once aggregated through ELITE, service enrollment becomes a proxy for enforcement targeting. The system doesn’t measure where deportable people are. It measures where the government already has the most data — which tends to be low-income communities, communities of color, communities that most rely on public services.
When ELITE ingests benefit program data and generates a confidence score, it is not measuring the probability that a deportable individual lives at a given address. It is measuring the depth of government surveillance of a given neighborhood, then treating that depth as evidence of enforcement potential. The methodology mistakes data availability for targeting validity.
COMPAS — the criminal risk assessment tool examined by ProPublica in 2016 — demonstrates the pattern. The investigation found that Black defendants were nearly twice as likely to be falsely flagged as high risk compared to white defendants. The algorithm was trained on historical data that encoded existing racial disparities. The model learned the bias and operationalized it as prediction. Garbage in, garbage out.
ELITE operates on similar logic. It doesn’t start with neutral data. It starts with decades of administrative practice that concentrated data collection in specific communities. The confidence scores reflect that history. A pin in a low-income neighborhood might have a high confidence score not because the person is more likely to be deportable, but because that neighborhood is more likely to be represented in government databases. The system treats visibility in administrative data as evidence of enforcement priority.
ELITE has not been subjected to any public independent audit. There is no published methodology for how confidence scores are calculated, what weights are assigned to different data sources, or how accuracy is assessed. ICE has made no public commitment to independent evaluation. We don’t know whether the system accounts for this data bias, whether it attempts to correct for differential surveillance of communities, whether anyone involved in building it considered these questions.
What we know about the data inputs tells us the system will generate the densest pin clusters in the most surveilled communities — systematically communities of color, immigrant communities, low-income communities. The confidence score doesn’t tell you where deportable people are. It tells you where the government has been paying the closest attention.
According to the Transactional Records Access Clearinghouse, more than 73% of people held in ICE detention have no criminal conviction. According to a Department of Justice document, of the 607 people arrested during Operation Midway Blitz, only only 2.6% were deemed high public safety risks. The gap between those figures and the official framing of ELITE as targeted, precise enforcement is structural, not incidental.
The Democratic Deficit
ELITE is one instance of a broader pattern: democratic governments build large-scale surveillance and targeting infrastructure without formal legislative authorization, then deploy it against populations with limited political recourse.
In the United Kingdom, the Prevent programme operates a referral system in which public sector workers — teachers, doctors, social workers — are required to flag individuals believed to be at risk of radicalization. The logic resembles ELITE’s confidence scoring: aggregate data from disparate sources (a student’s essay, a doctor’s observation, a teacher’s concern), weight by source authority, produce a risk assessment. The programme has been criticized for disproportionately targeting Muslim communities, with studies finding that referrals reflect demographic anxiety rather than genuine security risk. A 2022 independent review recommended significant reforms. The government accepted some and rejected others. The architecture remains in place. No primary legislation governs the referral process or the data it generates.
The common architecture: a surveillance or identification system is built for an administratively neutral purpose — welfare delivery, security screening, case management. The system generates data that is then repurposed for enforcement. The communities most affected are those already most heavily monitored. And in no case does primary legislation govern the expansion into enforcement use — it happens through executive orders, administrative guidance, contract addenda.
The democratic oversight gap is functional, not incidental. These systems are built in the gap between formal legislative authority and administrative practice precisely because formal authority would require deliberation about whether the targeting approach should exist at all.
The Harder Question
The standard accountability response to ELITE is that the tool is biased, produces errors, sweeps up innocent people, relies on data that communities never consented to share. Legal challenges from groups like Innovation Law Lab and advocacy from the Electronic Frontier Foundation and Just Futures Law have focused on transparency, data protection, and constitutional implications. Fix the bias. Improve the accuracy. Subject the scoring algorithm to independent audit. These are reasonable demands. But they rest on an assumption that the problem is implementation, not design.
If ELITE’s confidence scores became perfectly accurate — if the system could reliably identify undocumented individuals at specific addresses, with zero false positives — what would we have achieved? A more efficient version of area-based targeting. The methodology would remain unchanged: identify neighborhoods where pins cluster, deploy agents, conduct sweeps, detain broadly. It targeted a neighborhood. A more accurate ELITE doesn’t change that operational logic. It just makes it more efficient.
And beneath that question is a deeper one: what is this system being used for? The median undocumented immigrant has been in the US for 16 years. Many have US citizen children. They have been part of building American society — they work, pay taxes, are embedded in communities. Some arrived as children and have no memory of any other home. The United States is their home in every meaningful sense except legal status. Like the administrative limbo of Assam’s 1.8 million Aadhaar enrollees, many undocumented people in the US exist in a gap between documented status and lived reality. Is finding and removing people based solely on documentation status — people who’ve built lives here, raised families here, have no criminal record — the kind of enforcement we want to perfect with surveillance technology? Should determinations about who belongs be delegated to algorithmic systems built by private companies?
The bias critique operates within a framework of procedural fairness: fix the algorithm, and the process becomes legitimate. But this assumes the process itself — area sweeps without individual warrants, targeting communities based on demographic density, detaining people who have never been charged with crimes — is compatible with democratic governance. In democracies, enforcement power is constrained by individual rights. You cannot be detained without probable cause specific to you. You cannot be searched without a warrant. You are innocent until proven guilty. These are constitutional limits on state power designed to prevent exactly what ELITE enables: the targeting of communities wholesale rather than the investigation of individuals.
This raises a deeper concern: what happens when we allow even more complex black boxes to make deportation decisions? ELITE’s confidence scores are at least theoretically auditable — they measure address reliability based on source authority and recency. But imagine a machine learning system that determines deportability by processing hundreds of variables through neural networks. Even the engineers who build such systems often cannot explain why the model flags one person over another. The algorithm might identify patterns in the training data that correlate documentation status with factors no one intended: spending patterns, social media behavior, healthcare utilization, employment history. We wouldn’t know what it’s weighting. We couldn’t challenge its logic.
And once we accept opaque algorithmic determinations of belonging, what prevents future systems from incorporating other characteristics? Political affiliation mapped onto neighborhoods? Religious practice cross-referenced with benefit enrollment? Associational patterns derived from social networks? The answer is: nothing technical prevents it. What prevents it is democratic oversight. Once you normalize “the algorithm says you don’t belong here” as sufficient justification for detention, you’ve eliminated the mechanism through which democratic societies constrain power. The black box becomes unchallengeable because its reasoning is unknowable.
COMPAS illustrates the trap. After ProPublica demonstrated racial bias in COMPAS sentencing scores, some jurisdictions tried to build fairer algorithms. But that response skips a deeper question: should algorithmic risk assessment be used in sentencing at all? If the methodology treats group characteristics as evidence of individual culpability, it violates basic principles of liberal criminal justice. The problem is using group-level statistics to make individual-level determinations of guilt or risk — and doing it through black boxes that make fundamental decisions about people’s rights without explaining why.
ELITE operates on the same logic. The methodology — area targeting, community sweeps, density-based lead generation — is not “targeted enforcement” in any legally coherent sense. It’s community-level profiling that treats neighborhood residence as grounds for detention. And what that methodology targets — people based solely on documentation status, many of whom have lived here for decades with no criminal record — raises questions about who decides belonging and whether we should delegate those decisions to algorithmic systems we cannot inspect or contest.
The harder question isn’t whether “fix the bias” is the right demand. It’s whether we’re asking the right question at all. Not “how do we make ELITE more accurate?” but “is this what we want to use algorithmic systems for?” If we only fight the bias critique, we risk legitimizing the methodology. We risk accepting that algorithmic efficiency can replace constitutional constraints on state power. That questions of belonging — whether someone gets to stay in the only home they’ve known — can be delegated to confidence scores and density maps.
I don’t know how to hold both of those at once — the practical fight to reduce algorithmic harm and the structural fight about whether the system should exist at all. But I think a healthy democracy has to. Because if we only fight the bias critique, we risk legitimizing both the methodology and what it’s being used for.
The methodology is the problem. What it’s being used for is the problem. The precedent that black boxes can determine who belongs is the problem.
Pins or People
When I watch Aliya Rahman’s testimony, what I see is the operational consequence of a system that converted the question “who should we target” into “where should we go.” The map flattened Aliya into density metrics, density metrics into enforcement potential, enforcement potential into the decision to deploy agents to their neighborhood.
The investigation I’ve laid out here is incomplete. I found no evidence of ELITE being subjected to public independent audit. No published methodology for how confidence scores are calculated, what weights are assigned to different data sources, how accuracy is assessed. No insight into how this technology will evolve — what additional data might be layered in, what new variables and weights might be added, whether future versions might use machine learning to identify patterns invisible even to ICE. The opacity isn’t accidental. It’s structural.
I don’t know the training data provenance — what specific datasets were used, in what proportions, reviewed by whom. I don’t know whether Palantir engineers tested ELITE’s outputs for bias, whether they ran simulations to understand which neighborhoods would be most heavily targeted, whether anyone asked what would happen to people like Aliya Rahman. I don’t know whether ICE leadership reviewed the methodology before deployment, whether legal counsel assessed its constitutional implications, whether anyone considered that a system optimized for efficiency might violate fundamental principles of individualized suspicion. These questions matter. The absence of answers matters more.
What I have is sufficient to say this: the gap between what ICE says ELITE does and what agents describe using it for is not a gap of imprecision. It’s a gap of intent. The official language describes a tool that serves legal processes. The operational language describes a tool that selects communities.
In February 2026, a federal judge in Oregon ruled in favor of Innovation Law Lab, blocking ICE from making warrantless “dragnet-style” arrests in the state. The ruling came directly from legal challenges to operations like Black Rose. It’s one of the first judicial acknowledgments that algorithmic community targeting may be constitutionally incompatible with the individualized suspicion requirement that constrains enforcement power in democracies.
One ruling. One state. Against a surveillance apparatus spanning the country, built by a company with billions in government contracts, operating under legal authorities administratively adjusted to enable exactly this targeting.
The question isn’t whether the algorithm is biased. The question is what we’ve decided it’s for.
I’m Muskan, a data scientist who helps Series A-C startups build analytics infrastructure that drives decisions, not just dashboards. If you’re dealing with messy data, unclear metrics, or “we have data but can’t act on it” problems, let’s talk → [Calendly link]
Sources: Court transcript from Operation Black Rose proceedings (December 2025), purchased by 404 Media; Palantir contract modification documentation; reporting by 404 Media, Wired, ProPublica, Electronic Frontier Foundation, American Immigration Council, National Immigration Law Center, OpenDemocracy, MIT Media Lab Gender Shades, TRAC Immigration, UK Government Prevent Guidance, The Guardian, Al Jazeera, BBC, Pew Research, and Oxford Law Blogs; Congressional testimony of Aliya Rahman; ImmigrationOS contract documentation (Federal Contract ID: 70CTD022FR0000170).