Run: 2026-04-25 | Cluster: Linode g2-gpu-rtx4000a4-s (4× RTX 4000 Ada, us-ord) at 172.237.146.196
Pipeline: seed → recursive Serper deep-research → Gemma-4-E4B enrichment → NER round 1 → graph_sync
Executive summary
Nym Health (id=434009) is an autonomous medical coding company founded in 2018 in Tel Aviv by Amihai Neiderman (CEO) and Adam Rimon (CTO) — both alumni of IDF Unit 8200. A third co-founder, Yaron Spiegel, is in the graph but has minimal public footprint (Gemma honestly returned "no factual info available" rather than fabricate, then NER still extracted the founder edge from secondary sources). Nym has raised ~$92M total since founding (most recent: $47M in October 2024, prior $25M in July 2021, $16.5M Series A in October 2020). The company has been featured on TIME's top HealthTech list and positions itself as an "autonomous medical coding engine" that converts clinical narratives into ICD-10/CPT codes — directly contesting Abridge, Codametrix, Fathom, AKASA, and Tennr in the broader US medical-coding AI market.
Run metrics
- Targets pinned: 104 (Nym + 78 1-hop neighbors + 31 medical-coding peers/founders)
- Recursive Serper deep-research dossiers synthesized: 17/17 priority entities (avg 870 chars/dossier)
- Serper calls used in deep-research phase: ~150 (within budget; rate-limited to 3 s/call)
- NER round 1: 97 enriched entities processed in 24 min
- New entities created (system-wide): 94
- New relationships created (system-wide): 553
- Fresh edges directly on Nym + 3 founders: 99 total / 43 from NER round 1
- Fresh edges across the full Nym + medcoding peer set: 155 across 25 typed relationship types
- Approx spend: ~$32 of GPU + ~$1 of Serper
1. Nym Health profile (verified facts)
Company: Nym (canonical id=434009 in graph; also a NER-extracted variant Nym Health exists and should be merged via entity_alias_merge.py before round 2).
Founded: 2018, Tel Aviv, Israel. US presence in New York.
Product: Autonomous medical coding engine — clinical-language understanding that reads patient charts (notes, labs, imaging reports) and emits ICD-10/CPT codes without human review. Aimed at hospital RCM and physician-group revenue cycle.
Funding (graph-extracted):
- 2020-10-14: $16.5M Series A
- 2021-07-27: $25M (later round)
- 2024-10-02: $47M growth round
- Cumulative ≈ $92M Strategy edge (auto-emitted): Nym → adopts_strategy → "Focus on Specific Medical Specialties Drives Efficiency" — i.e., depth-first per-specialty taxonomy rather than horizontal coverage. Risk edge (auto-emitted): Nym → at_risk_from → "Accuracy Rate Decline" — the central durability question for any autonomous-coding vendor (a single high-profile under-coding incident at a hospital anchor client could unwind several quarters of growth). Opportunity edge (auto-emitted): "Expansion into New EMR Systems" → opportunity_for → Nym — Nym's growth lever vs. the ambient-AI scribe set is broader EHR coverage. Currently the graph shows Nym mentioned alongside Electronic Medical Records as a topic but no specific EHR integrations are recorded — a notable gap vs. peers (see §3). Hypothesis edge: "AI Accuracy Driven by Data Diversity" — relationship recorded both ways; the graph treats this as a research hypothesis worth testing against Nym's specialty-specific approach.
2. Founder network: Unit 8200 cluster
Amihai Neiderman (id=1917970, CEO)
- founded → Nym Health
- employed_by → Argus Cyber Security (his pre-Nym career; Argus was acquired by Continental for $400M+ in 2017 and is the prototype "Unit 8200 alum founds dual-use startup" pattern)
- profile_mentions → Unit 8200, Medical Coding, United States
- Background note (from synthesis): Nym was reportedly catalyzed by his wife's experience as an MD-PhD ENT surgeon — i.e., a domain-insider co-founder narrative that signals product-market understanding. Adam Rimon (id=1917971, CTO)
- founded → Nym Health (and
leadsedge as well) - related_to → Israeli Intelligence Corps (Unit 8200 sibling node)
- Met Neiderman in Unit 8200 (per dossier) — this is the standard Israeli-cyber-to-healthcare conversion pipeline. Yaron Spiegel (id=1917972, co-founder)
- founded → Nym Health
- Limited public footprint — likely earlier-stage technical co-founder; warrants targeted second-pass research on private filings. Pattern recognized: Two of three founders are Unit 8200 alumni; one had a prior cyber exit (Argus). This is the same founder-archetype pattern as Wiz, SentinelOne, and Cyera — and predicts: (a) defensive culture around enterprise security/compliance, (b) Israeli engineering core with US commercial leadership, (c) higher likelihood of strategic-acquirer exit vs. IPO at the 5-7 year mark.
3. Competitive landscape — what NER pulled
The graph now contains rich operating detail on every peer. CEO/leadership and integration partners that NER auto-extracted from web dossiers:
| Company | CEO (NER-extracted) | Key integration / partner edges | Specialization |
|---|---|---|---|
| Abridge | Shiv Rao | collaborates_with JAMA Network, NEJM | Ambient AI documentation |
| Tennr | Trey Holterman (corrected: my seed had "Lauderdale") | located_in New York, NY | Healthcare Operations / referrals |
| Suki AI | Punit Soni (corrected from seed) | integrates_with Epic EHR + MEDITECH Expanse; automates_coding_for ICD-10, CPT | Voice ambient documentation |
| DeepScribe | Matthew Ko | integrates_with Cerner + Epic | Ambient scribe |
| Fathom Health | Andrew Lockhart (corrected from seed) | specializes_in Autonomous Coding |
- Abridge → at_risk_from → "Vendor Dependency"
- Abridge → faced_issue → "Illegal Recording Of Patient-Clinician Conversations" (the real-world ambient-AI privacy risk that the entire scribe sub-segment shares)
4. Where Nym sits relative to peers
Nym's structural advantage:
- Pure autonomous-coding play (vs. Abridge/Suki/DeepScribe which are scribe-first and bolt coding on)
- Israeli engineering culture + Unit 8200 founders → defensive moat on data security and HIPAA posture
- $92M cumulative is mid-pack for the segment — Abridge has raised ~$460M, AKASA has raised ~$205M, Tennr is at ~$70M, Hippocratic AI is at ~$278M+. So Nym is well-capitalized for a focused product but not at the "horizontal platform" scale Abridge has chosen. Nym's structural gap (visible from the graph):
- Zero recorded EHR integration edges in our refreshed dossier vs. DeepScribe (Epic, Cerner) and Suki (Epic, MEDITECH). This is precisely what the auto-emitted "Expansion into New EMR Systems → opportunity_for → Nym" edge picked up.
- Zero recorded customer relationships — by contrast Codametrix's "demonstrated_effectiveness_at Mass General Brigham" edge is concrete proof; Nym needs equivalent named-customer edges to drive enterprise sales credibility. The bear case (also from the graph):
- "Accuracy Rate Decline" risk is the existential threat for a hands-off coding vendor. Once a hospital sees a CMS audit failure traced to autonomous codes, the procurement pendulum swings back to "human-in-the-loop" vendors (Codametrix, Iodine Software).
- "AI Accuracy Driven by Data Diversity" hypothesis (auto-emitted) cuts against Nym's specialty-first strategy: vendors with broader specialty coverage may, if the hypothesis holds, accumulate more accurate models faster.
5. Recommended follow-on research
- Canonicalize Nym (434009) ↔ "Nym Health" duplicate via
entity_alias_merge.pyso allfounded_byandraised_funding_amountedges from NER point at the canonical node before NER round 2. - Targeted second pass on Nym's investor list (the dossier mentions raises but no investor entities). Likely cap-table participants based on founder pedigree + round size: Lightspeed, Bessemer, Addition, Tiger, Adams Street, Israeli growth-stage funds (Vintage, Bessemer Israel). Pin those entities and run NER round 2.
- Customer / hospital edges for Nym — currently zero in graph; this is the most valuable next pull. Run a Serper sub-query:
"Nym Health" hospital OR health system OR customerand re-NER the resulting dossier. - Run the same recursive Serper deep-research script on the still-never-enriched ~50 1-hop neighbors of Nym (~$2 Serper, ~5 min wall-clock).
- Comparative funding edges —
raised_funding_amountedges currently exist only for Nym; backfill for all peers to make the/companiesfinancial-comparison view useful.
6. Cluster status & infrastructure
- Linode
panopticon-nym-medcoding-202604250522(id 96621161) at172.237.146.196— still healthy. - All 8 llama-server processes (4× Gemma-4-E4B :8000-:8003 + 4× Jina v5 :8004-:8007) running.
- Background
enrich_cli.py loop --priority top --top 30 --interval 180still cycling. graph_sync.py --incrementalrunning on the burst every 5 min (also being run from local for the canonical Neo4j push).- Spend so far: ~$32 (≈9 h × $3.60/hr).
7. Knowledge graph export to Pano
7.1 Run totals (rounds 1 + 2 combined)
- Targets pinned: 104. Recursive Serper deep-research dossiers synthesized: 17/17.
- NER round 1: 24 min, 94 new entities, 553 new relationships (333 sub-entities surfaced).
- NER round 2: 27 min, 21 new entities, 569 new relationships (332 sub-entities surfaced).
- Combined: 115 new entities + 1,122 new relationships added to Supabase + Neo4j.
- 725 fresh Entity nodes in Neo4j (since 2026-04-25 05:00 UTC).
- 22 fresh Article nodes, 4,620 fresh
RELATES_TOedges, 497 freshMENTIONSedges synced viagraph_sync.py --incremental(delta includes both targeted-burst output and background loop output).
7.2 Dashboard verification (Pano /entities, /companies, /graph)
All 21 priority entities now have: (a) refreshed description, (b) deep-research background (300–1,400 chars), (c) non-zero edge degree, (d) Jina-v5 1024-dim embeddings.
Degree of priority entities in Neo4j after both rounds:
- Nym (434009): 68 edges
- Abridge: 30, AKASA: 28, Augmedix: 32, DeepScribe: 29, Fathom Health: 22, Suki AI: 22, Iodine Software: 21, Codametrix: 19, Tennr: 7, Hippocratic AI: 4, Waystar: 11, XpertDox: 11, Notable Health: 8
- Amihai Neiderman: 23, Adam Rimon: 25, Yaron Spiegel: 4 (matches low public footprint), Shiv Rao: 19, Trey Lauderdale: 16, Munjal Shah: 14, Malinka Walaliyadde: 0 (NER missed her dossier; needs targeted re-run)
7.3 Visualization issues to fix
- Entity 434009 "Nym" is conflated between Nym Health (autonomous medical coding) and NymVPN (the privacy/cybersecurity project). The
/entities/434009view shows mixed edges:raised_funding_amount → $47 Million(Nym Health) sitting next todeveloped → NymVPN,focus_on → Cybersecurity,focus_on → Privacy infrastructure. This is not a duplicate problem — it's an over-merge problem that requires a SPLIT operation, notentity_alias_merge.py. Recommended fix: create a new entity "Nym Health" with slugnym-health, redirect the medical-coding edges (founders, $47M raise, autonomous coding) to it, and leave 434009 as the VPN entity. - Cluster cohesion is weak. Only 8/20 priority entities are reachable within 2 hops of Nym. Direct path queries
Nym → Abridge,Nym → AKASA,Nym → Codametrix,Nym → Fathom Healthall return "not connected within 2 hops". Each company is well-connected to its own context (founders, locations, products) but the peer-to-peer competitive edges are missing because no single dossier cross-references its rivals. Recommended fix: run a Serper-driven "competitive landscape" extraction pass ("<peer A>" vs "<peer B>" medical coding) that explicitly producescompetitor_ofedges between the 13 peers. - Tennr (1586258) and Hippocratic AI (1586257) were missing from Neo4j after the initial graph_sync (they pre-existed in Supabase but only got new edges during the run). Patched in via
MERGEin this session; document going forward that should also push edge-only deltas for pre-existing nodes.
7.4 Pano dashboard pages now live with this run
/entities/434009— Nym dossier (mixed Nym Health + NymVPN; see 7.3.1)/entities/1917970(Amihai Neiderman, 23 edges) and/entities/1917971(Adam Rimon, 25 edges) — full bios with Unit 8200 / Argus /founded/leadsedges/entities/1917961through/entities/1917969— the 9 newly-seeded medical-coding competitors with multi-paragraph dossiers/companies/abridge,/companies/waystar,/companies/r1-rcm— refreshed comparison views (slug-keyed; existing rows from prior runs)/graph— medical-coding subcluster, 17,249 nodes reachable within 2 hops of Nym/search(semantic) — Jina-v5 1024-dim embeddings on all 17 deep-researched entities
8. Resolution: entity-split + cluster-cohesion patch (2026-04-25 17:50 UTC)
8.1 Entity disambiguation
Discovered that a canonical Nym Health entity already existed at id=561480 in Supabase (created during a prior ingest). Promoted 561480 to canonical and:
- Categorized the 69 edges on 434009 into 29 health / 25 VPN / 15 unknown.
- Redirected all 29 medical-coding edges from 434009 → 561480 (founders, $47M raise, RIOH edges, TIME recognition, autonomous-coding mentions).
- Wrote the deep-research dossier (≈1.4 KB description + 643 c background) to 561480.
- Linked 561480 to environment 40 (medical-coding watchlist) and pinned
always_enrich=true. - Left 434009 as the clean
Nym/ NymVPN privacy-cybersecurity entity (68 remaining edges, all VPN-coherent).
8.2 Cluster-cohesion densification
Added 18 explicit competitor_of edges + 15 cross-cohesion edges between the 13 medical-coding peers (autonomous-coding cohort, ambient-scribe cohort, RCM cohort), bidirectional. Total 33 unique pairs × 2 directions → 66 + 15 = 81 net new typed edges added to Supabase, all synced to Neo4j.
8.3 Final Pano dashboard state (verified)
- Cluster cohesion: 20/20 priority entities now reachable within 2 hops of Nym Health (was 8/20 before patch).
- Nym Health (561480): 106 edges in Neo4j; rich dossier; direct
competitor_ofto Abridge, AKASA, Codametrix, Fathom Health, Augmedix, Iodine Software, Suki AI, DeepScribe, XpertDox (all hops=1). - Nym VPN (434009): 68 edges, all VPN-domain coherent.
- Intra-cluster density: 228 edges within the 21-entity priority set.
- Per-entity degree (Neo4j, post-patch): Nym Health 106, Abridge 50, AKASA 45, Augmedix 40, DeepScribe 37, Iodine 35, Fathom 34, Codametrix 31, Suki AI 30, Adam Rimon 25, XpertDox 25, Amihai Neiderman 23, Shiv Rao 19, Waystar 17, Trey Lauderdale 16, Munjal Shah 14, Notable Health 12, Tennr 11, Hippocratic AI 8, Malinka Walaliyadde 5, Yaron Spiegel 4.
- Pano dashboard URLs now coherent:
/entities/561480— canonical Nym Health dossier/entities/434009— Nym VPN, no more health-coding pollution/graph— connected medical-coding subcluster, 228 intra-cluster edges
9. Remaining backlog (post-resolution)
- Targeted second pass on Nym Health's investor list (Lightspeed, Bessemer, Addition, Tiger, Israeli growth funds) — still 0 investor edges in graph.
- Customer / hospital edges for Nym Health — still 0 in graph; Serper sub-query
"Nym Health" hospital OR health system OR customer. - Run recursive Serper deep-research on the still-never-enriched ~50 Nym 1-hop neighbors (~$2 Serper, ~5 min wall-clock).
- Patch
graph_sync.pyto UPDATEdescription/background/enrichment_scoreon EXISTING Neo4j nodes (currently only INSERTs new). Workaround in this session: direct CypherSETcalls. - Add Nym Health to
frontend/src/lib/company-library.tsandcompany-environments.tsso it gets a first-class/companies/nym-healthpage (currently only renders via/entities/561480). Co-Authored-By: Oz oz-agent@warp.dev