Key text here may be On opinion, knowledge and belief, CPR B 848-859.
There is conviction [Überzeugung]. It is the subjective part necessary for knowledge:
Taking something to be true is an occurrence in our understanding that
may rest on objective grounds, but that also requires subjective causes in
the mind of him who judges. If it is valid for everyone merely as long as
he has reason, then its ground is objectively sufficient, and in that case
taking something to be true is called conviction. (B848)
The "objective" aspect of knowledge is "certainty" [Gewissheit], and only if both occur together, this can be called knowledge. Kant speaks of three possible variations in respect to conviction:
Taking something to be true, or the subjective validity of judgment,
has the following three stages in relation to conviction (which at the
same time is valid objectively): having an opinion, believing, and
knowing. Having an opinion is taking something to be true with the
consciousness that it is subjectively as well as objectively insuffıcient. If
taking something to be true is only subjectively sufficient and is at the
same time held to be objectively insufﬁcient, then it is called believing.
Finally, when taking something to be true is both subjectively and ob-
jectively sufﬁcient it is called knowing. Subjective sufﬁciency is called
conviction (for myself), objective sufﬁciency, certainty (for everyone). (B850)
Opinion: We are neither convinced regarding the objectivity of the proposition because of a valid judgement nor certain that this judgement will be the same for every rational being.
Belief: We are convinced that it is true because it is the result of valid reasoning based of our personal experience ("subjective causes in the mind of him who judges"), but not certain that it is the same for every rational being.
Knowledge: We are convinced that it is true because it is the result of valid reasoning based on our personal experience ("subjective causes in the mind of him who judges"), and certain that it is the same for every rational being (because even if it is based on personal experience (empirical), it is - itself - not reliant on that particular experience).
It is different from JTB in that beliefs - taken this to be true - are always justified. Also, "truth" cannot be empirical truth, but has to be a priori, as only there, certainty is to be found.
To further bolster my assessment as being backed by Kant's own words:
l must never undertake to have an opinion without at least knowing something by means of which the in itself merely problematic judgment acquires a connection with truth which, although it is not
complete, is nevertheless more than an arbitrary invention. Furthermore, the law of such a connection must be certain. For if in regard to
this too I have nothing but opinion, then it is all only a game of imagination without the least relation to truth. In judging from pure reason,
to have an opinion is not allowed at all. For since it will not be supported on grounds of experience, but everything that is necessary
should be cognized a priori, the principle of connection requires universality and necessity, thus complete certainty, otherwise no guidance
to the truth is forthcoming at all. (B850-51)
One aspect is that of "problematical" judgements, which could be understood as "may be"-judgements (see e.g. GMM, 4:414-5). The other one is that he seems to claim here that all knowledge (other than a priori principles themselves) has to be derived from a priori principles, which, albeit "resting on empirical grounds", remain a priori (see Prol., Ak. 4:373-4,fn.). This is closely linked to Kant's understanding of science, which I have elaborated in the context of this answer.
On truth, a priori, and certainty
The status of a priori knowledge is questionable and I am sympathetic with a reading originally worked out by Onora O'Neill ('89, '90, '92) and summarised by Pawel Lukow (Kant-Studien 84:2, 204-221, pp. 205-6) as follows:
According to a constructivist reading of Kant's approach, philosophy uses and
can only use a negative procedure: it rejects all claims that do not survive self-
criticism, and all claims that cannot be followed by all. Even the process of vindicating reason uses, and can only use, these two strategies. On the one hand
we reject principles which, as far as we can determine, are self-defeating, and, on
the other hand, we reject principles that not all can follow. In consequence, we
provisionally retain only principles which do not contradict themselves and which
we hold that all can follow. By these procedures, we arrive at an account of certain
principles — which we come to view as those of reason — which resist refutation
when subjected to reflexive self-criticism. This is not to say that these principles
are the right ones, judged by some transcendent conception of reason, or that these
are all the unrestricted legitimate principles there are. The principles that survive
testing are still open to criticism. Vindication of reason is viewed as an incompletable
task, standards of reasoning are provisional: construction is a matter of (negatively)
rejecting some provisional proposals and retaining others rather than that of
(positively) proving determinate proposals.
This view of vindication is exemplified by Kant's specifically juridical6 under-
standing of 'deduction'. He makes a trial (or experiment; Versuch] where main
procedure is negative. A 'deduction' is not a logical proof but legitimization
(Rechtfertigung-, cf. for example Kr.d.p. V., 45).
Thus, arguably, a "deduction" (juridical sense) that shows that a proposition is not self-contradictory and can be followed by all is what makes a belief justified in the sense of certain beyond a reasonable (rational!) doubt. Mind, for Kant reason works on the basis of experience (Prolegomena, 4:373 fn.) and pure speculation is irrational and has to be vindicated by critical reason. In other words: Radical scepticism with Cartesian spectres like Brains-in-a-vat is irrational as in "purely speculative".
Also, there is no differentiation between true and justified because there would have to be an independent, justifiable, true criterion for truth leading to an infinite regress or scepticism.
Since the Kantian conception of knowledge is supposed to defeat sceptical objection above all else, it seems fitting that Lukow concludes:
Although the authority of reason cannot be proved in this way, it has
at least survived refutation. (207)
Altogether there are some similarities, but I would say the Kantian concept of knowledge is much more restrictive since his understanding of truth forbids knowledge being empirical - following Hume's philosophy but going beyond.
The main difference is that there is no differentiation between true and justified since the justification is what determines the truth.
Edit as I stumbled across it:
In his Metaphysical Foundations of Natural Sciences Kant restates this strict sense of knowledge, saying that knowledge [Wissen] that is founded in empirical findings is not to be called knowledge [Wissen] in a strict sense (Ak. 4:468). He distinguishes once more between knowledge [Erkenntnis], that is apodictically certain and systematically gained by principles a priori, and experience [Erfahrung] that is founded by empirical facts and/or gained through empirical principles.
What is therefore needed for knowledge is the "construction of the concept" (any Hegelians knowing this? now you know how strong Kant's influence was!) through schematism, i.e. especially pointing out the connection between a priori and intuition in order to prove the objective reality of the concept. Förster describes this in his chapter on the Foundations, reflecting on the preface.