×
Antfly
A search-and-inference database for AI agents: full-text, vector and graph indexes with the models running inside the engine
Last verified September 21, 2026 · Updated daily
What Antfly is building
Antfly is a database with the retrieval pipeline built in. A single table can carry full-text (BM25), dense, sparse and late-interaction vector indexes plus graph edges at the same time, and the models that chunk, embed, rerank and extract entities run inside the engine rather than behind API calls. Drop in PDFs, screenshots, tables or audio and it indexes, embeds and links them on ingest; query with one plan that fuses lexical, semantic and graph retrieval. The premise, in James McDermott's words: 'Every AI team eventually builds the same retrieval stack with the same glue code.' The engineering is the interesting part. The core is written in Zig with zero dependencies: a custom LSM storage engine, an LMDB-compatible B+ tree, a write-ahead log, Multi-Raft with separate consensus groups for metadata and storage, Lucene-style full-text segments, vector indexing with RaBitQ compression and SPFresh-style updates, and their own HTTP/2 and HTTP/3 and inference runtime. Critical distributed protocols have TLA+ specifications and there is deterministic VOPR-style fault-injection testing. It runs as one binary on a laptop, a Kubernetes cluster, an embedded .aflite file, WASM or via a C API, with CUDA and Metal acceleration, Postgres CDC and S3 storage. The core server is Elastic License 2.0; SDKs, React components, the inference runtime and tools like pgaf, docsaf and evalaf are Apache 2.0. Built-in models include CLIP for text and image, GLiNER2 for zero-shot entity and relation extraction, and CLAP for audio. First public customer is Visiting Media, a Portland hospitality sales platform, which indexed more than 100 terabytes of images, video and text in under 24 hours and built an agent that retrieves assets and assembles a sales proposal in under a minute.
Why this matters
The retrieval gap is the practical reason agents fail in production: the context an agent needs is spread across a vector store, a keyword index, a graph, a reranker and a sync pipeline, and each seam is a place for the wrong document to win. Joe Ruscio at Heavybit puts the investment thesis in one line: 'Antfly closes it with lexical, semantic, and graph retrieval in one framework, fine-grained access controls, and deployment in your own environment.' The deployment part matters as much as the search part: inference never leaves your infrastructure, which is the constraint enterprises actually have. The founders have done the data-platform version of this before. McDermott built Lytics, a customer data platform that ran at a scale of billions of profiles and was acquired by Contentstack, and Roetker built its data platform. Heavybit is the developer-infrastructure specialist fund (its portfolio is where a lot of the current observability and data tooling came from), so a Heavybit lead on a Zig database is a signal that the engine is real, not a wrapper.
Open roles at Antfly
3 positions we're tracking. Roles are re-checked daily and removed when filled.
Founding Engineer, storage and distributed systems (Zig; inferred)
First seen today
Founding Engineer, ML inference and retrieval quality (inferred)
First seen today
Developer Relations / Growth Engineer (inferred from the self-host-first model and Portland meetup activity)
First seen today
Know when Antfly is hiring before anyone else
A role stays uncontested for about four days. Here's the window — and where we put you in it.
From $5.99/week, cancel any time.
Watching Antfly
0 applicantsRole spotted & verified
1You get the alert
1You've applied
~8Hits the job boards
250+Hiring outlook
No hiring language in the announcement and no careers page; the stated use of funds is 'expand operations and its business development efforts'. The team page lists six people including three founding engineers and a head of growth, and the funding post invites teams with retrieval problems to get in touch. $2M at pre-seed with a shipping engine points to one or two very targeted hires, not a board.
Working at Antfly
Antfly is a AI company based in Portland, USA. Working at a AI company at this stage means opportunity to shape your role based on the company stage.
Portland, USA is where most Antfly positions are located.
How to actually get hired at Antfly
Why applying the normal way doesn't work
There's an ATS, but at Antfly, referrals get priority. Your cold application competes with sourced candidates and internal recommendations.
Who to contact at Antfly
What to show them
Clone github.com/antflydb/antfly, run the fault-injection tests, and either land a real PR (a bug in the LSM compaction path, a Raft edge case, a benchmark) or write a review of one subsystem against its TLA+ spec with a concrete divergence you found or a property you'd add. A merged PR beats any CV here.
A cold email that works at Antfly
What Antfly screens for
Heavybit and 8-Bit leading a $2M pre-seed for two ex-Lytics founders is a getting-the-band-back-together round: the Portland team, the same data-platform problem, one level down the stack. The tell for candidates is the licence split. ELv2 on the core and Apache 2.0 on everything around it means the business is Antfly Cloud, currently a waitlist, so the next hires are whoever makes the managed service run and whoever makes self-hosters convert. With 454 stars and one named customer, the company is pre-community; the person who builds the docs, benchmarks and meetup presence has outsized leverage.
Customize your CV for the Antfly role. Matching the job description language helps clear ATS filters.
Don't make these mistakes
James worked at WebTrends and founded Storycode and Lytics before this and has heard every 'RAG is broken' pitch; do not open with the retrieval gap, he wrote the press release about it. AJ built the data platform at Lytics and has been talking about agents being 'starved for the right context at the right time' at the Portland AI Engineers meetup since April; a pitch about vector databases generally will read as not having looked at the repo. The engine is Zig with no dependencies and custom Raft, LSM and B+ trees, so if you have never written a storage engine or a consensus protocol, lead with the retrieval-quality or inference side instead of pretending.
Mistakes that kill Antfly applications
A recycled CV gets rejected fast at Antfly ( people). They notice.
Don't open with what you want. Open with what Antfly is dealing with right now — The retrieval gap is the practical reason agents fail in production: the context an agent needs is spread across a vector store, a keyword index, a graph, a reranker and a sync pipeline, and each seam is a place for the wrong document to win, and what you'd do about it.
Most AI applications get ghosted. A day-five follow-up can double your response rate.
Applying to Antfly? Get the contact, not the form.
The Antfly interview process
4 stages · 14 days typical · take-home: yes · modelled from similar companies
We don't yet have verified candidate reports for Antfly. What follows is the typical process for a -person AI company — treat it as a model, not confirmed detail.
Interview stages
Recruiter Screen
Phone or video · 30 min
Basic qualification and logistics
Recruiter or HR
Hiring Manager Interview
Video call · 45 min
Role fit and experience deep-dive
Hiring manager
Technical/Functional Round
Video call · 60 min
Skills assessment and problem-solving
Team members
Final Round
In-person or video · 60 min
Culture fit and cross-functional alignment
Senior leadership
Antfly take-home assignment
Antfly includes a take-home exercise in their interview process. For AI roles, this typically involves a practical problem that takes 2-4 hours. Focus on clean, working code over premature optimization. They're evaluating how you think and communicate, not just the solution.
Antfly interview timeline
Expect days total. Compared to similar AI companies (14 days median), Antfly is about average.
Interviewed at Antfly?
Tell us how it went — stages, questions, timeline. Takes 90 seconds and it's how this page stays accurate for the next person.
Submit your Antfly interview experience →Antfly jobs, frequently asked questions
How many jobs does Antfly have open?
3 open roles at Antfly, last checked September 2026.
Does Antfly hire remotely?
Currently, Antfly only has in-office roles in Portland, USA.
What roles is Antfly hiring for?
Antfly is hiring across Engineering. The most recent opening is Founding Engineer, storage and distributed systems (Zig; inferred).
How do I apply for a job at Antfly?
Use the apply links above, or check our guide to getting hired at Antfly.
Does Antfly respond to cold emails?
Response rate data for Antfly not yet confirmed.
Who is the hiring manager at Antfly?
At this size, hiring is usually run by the hiring manager.
How competitive is it to get hired at Antfly?
Expect 100-250 applicants in the first two weeks for AI roles at this size. Apply within 72 hours for best odds.
How many rounds is the Antfly interview?
4 stages: Recruiter Screen, Hiring Manager Interview, Technical/Functional Round, Final Round.
Is the Antfly interview hard?
Focus on technical depth and system design, not puzzles. Technical Interview is reportedly the most challenging round.
Does Antfly give a take-home task?
Yes, Antfly includes a take-home assignment.
How long does Antfly take to get back to you?
Around 14 days across the full process.
What should I prepare for the Antfly interview?
Focus on technical depth and system design. At people, they're testing whether you can operate without process, not whether you memorised algorithms.
Where is Antfly based?
Antfly is headquartered in Portland, USA.
Get Antfly roles before they're posted
A role stays uncontested for about four days. Here's the window — and where we put you in it.
From $5.99/week, cancel any time.
Watching Antfly
0 applicantsRole spotted & verified
1You get the alert
1You've applied
~8Hits the job boards
250+Related
More AI companies hiring
Other startups we track in the same space.