Precision Federal is pursuing opportunities at DTRA
Precision Federal is pursuing opportunities at the Defense Threat Reduction Agency across SBIR Open Topics, the Research and Development Directorate's BAAs, and direct awards. DTRA is the Department of Defense's combat support agency for countering weapons of mass destruction and emerging threats — chemical, biological, radiological, nuclear, and high-yield explosive. Headquartered at Fort Belvoir, Virginia, with a workforce of roughly 1,400 DoD civilians and 800 uniformed service members across more than a dozen worldwide locations, DTRA carries a multi-billion-dollar mission portfolio with growing software, AI, and data-fusion demand inside that envelope.
Precision Delivery Federal LLC (UEI Y2JVCZXT9HP5, CAGE 1AYQ0, NAICS 541512) is SAM-registered, DSIP-active, and structured to engage DTRA directly on SBIR efforts, with partner firms joining larger CWMD performer teams. Led by a Kaggle Top 200 data scientist with prior federal delivery on production federal ML at a federal health agency, we bring the combination DTRA performer teams need: someone who can both build research-grade fusion models and ship them through security review.
How DTRA contracting actually works
DTRA is a combat support agency under the Office of the Under Secretary of Defense for Research and Engineering. Its contracting differs from typical service-component buying in three ways:
- R&D Directorate is the primary technology buyer — DTRA's Research and Development Directorate executes the bulk of CWMD technology investment, with mission scope spanning prevention, detection, attribution, response, and recovery across CBRN.
- SBIR Open Topics are the small-business on-ramp — DTRA participates in the DoD SBIR cycles via DSIP. Topics are typically Phase I-only with award targets in the ~$250K / 6 month range, and Open Topics in particular invite non-traditional, software-first solutions to long-standing CWMD coverage gaps.
- The CWMD R&D Directorate Support IDIQ — a 10-year multiple-award IDIQ with a $4 billion ceiling awarded in late 2024 to nine prime performers across three pools: AI/ML/data science/software (Pool 1), CBRN operations and countermeasures (Pool 2), and targeting/information operations/irregular warfare (Pool 3). For small businesses, this is the Phase III transition target.
DTRA SBIR topics carry SITIS Q&A windows during the open period (unlike DARPA), giving proposers a structured way to pre-clarify scope. Many DTRA topics are ITAR-controlled — proposers must be prepared to file evidence of DD Form 2345 (Militarily Critical Technical Data Agreement) submission inside the proposal package per the 2026 BAA.
DTRA mission scopes we target
Radiological / Nuclear detection from non-bespoke signals
Our primary target. DTRA's CWMD coverage problem is economic — bespoke R/N sensors do not scale to match proliferating threat volume. Our agentic AI and ML practice builds multi-modal fusion across already-deployed military and commercial sensors plus public data sources, producing geo-tagged threat indicators with provenance and uncertainty quantification.
CBRN threat anticipation analytics
Pattern recognition across logistics, OSINT, environmental, and behavioral signals to indicate WMD storage, transfer, and use. ML on the same kind of multi-source, multi-classification federal data streams we have shipped before — translated into the CWMD context.
CWMD data engineering and provenance
Lakehouse architectures with provenance attribution, reproducible pipelines, and uncertainty quantification — accepted high-stakes-intel practice. See Data Engineering.
Adversarial ML hardening
CWMD models will face adversarial conditions by definition. Robustness against evasion, poisoning, and prompt-injection attacks is part of the product, not an afterthought.
Our DTRA proposal: DTRA254-P005
Precision Federal submitted a Phase I proposal to DTRA254-P005 — Novel Technologies for CWMD and Related Threats (Open Topic) in the DoD 25.4 Release 12 cycle in May 2026. The topic invites software and analytics solutions that detect the storage, transfer, or use of WMD using signals from already-deployed general-purpose military hardware, non-specialized commercial equipment, and publicly-available data sources — explicitly without bespoke R/N sensors. That framing maps directly onto our multi-modal OSINT-fusion approach.
- Non-bespoke sensor signals — microphones, cameras, motion detectors, passive infrared, and other common sensors already deployed across military platforms and consumer devices, fused with public environmental baselines.
- Multi-modal data fusion — late-fusion transformer architectures combining per-modality anomaly detectors into geo-tagged R/N indicator scores. AIS, ADS-B, commercial EO thermal imagery, traffic cameras, OSINT, and environmental baselines as feeder modalities.
- Provenance and uncertainty quantification — every score carries source attribution and a calibrated confidence band, matching how high-stakes intelligence consumers actually need to act on a model output.
- Dual-use commercial framing — port-authority threat intelligence, supply-chain risk analytics, and insurance/reinsurance risk modeling. The topic explicitly requests dual-use applications.
- Phase I deliverable: proof-of-study — feasibility validated on historical and synthetic scenarios, with a Phase II course of action toward operational prototype.
SBIR Open Topics — the small-business on-ramp
DTRA Open Topic SBIRs are arguably the single most accessible DTRA vehicle for a SAM-registered small business with strong AI capability. Defining features:
- Award target: Phase I in the ~$250K / 6 month range (verify per BAA release).
- Phase II follow-on for selected performers, scaling award and timeline materially.
- Phase III transition path into the DTRA CWMD R&D Directorate Support IDIQ task-order ecosystem or DHS / CBP / NNSA parallels.
- SITIS Q&A windows during the open period — proposers can pre-clarify scope before the technical volume freezes.
- Many topics are ITAR-controlled — DD Form 2345 evidence of submission belongs in the proposal package.
We monitor open DTRA SBIR releases continuously and pursue those where our capability stack — multi-modal ML fusion, federal deployment discipline, OSINT data engineering — matches the topic's intent. Our differentiator is federal production experience, not academic publication record; that means we position well on Open Topics where the program owner wants a proof-of-study that survives beyond the lab demo.
Capability map for DTRA scope
Multi-modal ML fusion / agentic AI
Late-fusion transformer architectures, per-modality anomaly detectors, deep-ensemble uncertainty quantification, geo-tagged threat indicator output. See Agentic AI.
Machine learning and data science
Anomaly detection, time-series forecasting, CV on EO/thermal imagery, trajectory analysis on AIS/ADS-B, NLP on OSINT corpora.
Cloud architecture, IL4/IL5
Azure Government, AWS GovCloud, on-premise air-gapped. ATO-minded from day one.
Data engineering with provenance
Lakehouses with provenance attribution, reproducible pipelines, governed analytics, multi-classification handling. See Data Engineering.
Adversarial ML and DevSecOps
STIG compliance, NIST 800-171 self-assessment, CMMC L1 today / L2 plan for Phase II, prompt-injection hardening for any LLM-in-the-loop scope.
Federal delivery record
Precision Federal's own federal contracts are FHWA purchase order 693JJ326P000024 ($141,370.26, September 2026 through September 2028) and CDC purchase order 75D30126P21398 ($296,250, signed September 18, 2026). Before founding the firm, Bo Peng delivered this federal work at prior employers:
- Federal health agency ML — production ML system on federal health data, through full ATO, real users (prior federal delivery).
- Federal health IT data platform — lakehouse architecture, governance-first (prior federal delivery).
- Multi-agency cloud migration — delivered through federal contractor engagements (prior federal delivery).
For DTRA specifically we are targeting and pursuing through SBIR Open Topics and CWMD R&D Directorate BAAs. Why this still matters to DTRA program owners: the biggest gap on CWMD performer teams is usually not novel research — it is the engineer who can take the research prototype through security review, deployment, and operational hand-off. That engineer is us.
Vehicles, NAICS, and engagement
- Primary NAICS 541512 — Computer Systems Design Services. SBA small business.
- Additional NAICS on SAM — 541511, 541519, 541715, 518210.
- DTRA vehicles we're positioned for — DTRA SBIR Open Topics on DSIP and CWMD R&D Directorate BAAs, with partner firms on multi-performer CWMD teams.
If you are a DTRA program owner or a university PI building a CWMD team, email contact@precisionfederal.com. See also SBIR partnering, How We Work and Partners.