OpenAI powered next level of bench exchange and staff augmentation
EMBIQ's own R&D initiative from 2023: an internal assistant that uses the OpenAI API to match anonymised specialist profiles with staff augmentation and bench exchange requests. It was built and piloted in-house as a prototype, alongside a longer-term concept for machine-readable developer and project profiles.
Technologies used
Python, Django, PostgreSQL, Angular
Tools
OpenAI API, ChatGPT, Slack, Zapier, Bitbucket
4
Project members
7 months
Project length
AI Development, Custom Software Development
Category
HR
Branch
.png)
The challenge
Like many software houses, EMBIQ regularly has specialists between projects and, at the same time, receives requests for people it cannot staff itself. In early 2023 we set out to make exchanging such capacity with partner companies faster, starting with bench exchange outreach to software houses in Poland, Austria and the wider DACH region. The practical bottleneck was matching: requests arrive as free text through many different channels, while candidate profiles are descriptive documents, so comparing the two was manual, slow and subjective. This was an internal project, not a client commission, and the requirements came from our own sales and HR teams. The assistant had to know every profile held in our existing anonymised profile generator, including several variants of one person's profile; pick the best-fitting people for a request, and different people when a request covers several roles; and say plainly when nothing fits instead of forcing a match. Candidates' identities had to stay protected throughout, so the language model was only ever allowed to work on profiles that had already been anonymised.
Working on this project was a unique experience. Integrating OpenAI with our system brought a new level of precision and efficiency. It's amazing to see how technology can transform the recruitment process, making it more streamlined and unbiased.
Key Features
- Matching of free-text requests against the full base of anonymised specialist profiles
- Awareness of several profile variants for the same specialist
- Selection of different candidates for requests that cover several roles
- A clear "no match" answer when no suitable profile exists
- Automatic intake of requests from a partner community channel, staffing agency e-mails and the general inbox
- Slack bot that brings the most relevant requests and suggestions to the sales team
- Language model applied only to data that has already been anonymised
.png)
The result
By summer 2023 the prototype was working on real traffic. From July, requests posted in a partner community's bench exchange channel and sent by staffing agencies were forwarded automatically to a dedicated mailbox, and a Slack bot built on the OpenAI API and connected to our profile generator was used to process them. The team tuned the model's behaviour in short iterations, including limits on response length, and by September 2023 AI-assisted matching of candidates to requests had become part of how EMBIQ handles staffing enquiries. The prototype also gave us a concrete basis for a longer-term concept: machine-readable developer profiles derived from source code repositories, matching "need" profiles generated from a project's repositories, and anonymous matchmaking in which contact is established only when both sides confirm interest. That concept has not gone beyond the planning stage; what was built and used is the internal matching assistant described above.
- Requests from several channels collected automatically in one place
- First shortlist of anonymised profiles prepared by the assistant instead of by hand
- Candidate identities kept out of the language model
- Hands-on experience with integrating a large language model into everyday internal workflows
The collaboration between our team and OpenAI has been seamless. The challenge was not just about integrating a technology but ensuring that it aligns with our core values of data privacy and unbiased talent matching. I'm proud of what we've achieved.
What kind of competencies are you looking for?
Contact me to discuss your requirements and build your dedicated IT team together with us.
Agnes
business development specialist
Get a quote or advice
Let’s investigate your project concept and its current status together.
Expect an initial project scope proposal, time and cost estimation from us.
The consultancy will be protected by the NDA.